siglab-py 0.5.66__py3-none-any.whl → 0.6.18__py3-none-any.whl

siglab_py/backtests/backtest_core.py

@@ -0,0 +1,2371 @@
+# type: ignore Sorry sorry
+import os
+import logging
+import argparse
+import arrow
+from datetime import datetime, timedelta, timezone
+import time
+from typing import List, Dict, Any, Union, Callable
+import uuid
+import math
+import json
+import inspect
+import pandas as pd
+import matplotlib.pyplot as plt
+import matplotlib.dates as mdates
+
+from ccxt.base.exchange import Exchange
+
+from siglab_py.util.retry_util import retry
+from siglab_py.util.market_data_util import fetch_candles, fix_column_types, fetch_historical_price, timestamp_to_week_of_month
+from siglab_py.util.trading_util import calc_eff_trailing_sl
+from siglab_py.util.analytic_util import compute_candles_stats, lookup_fib_target, partition_sliding_window
+from siglab_py.util.simple_math import bucket_series, bucketize_val
+
+def get_logger(report_name : str):
+    logging.Formatter.converter = time.gmtime
+    logger = logging.getLogger(report_name)
+    log_level = logging.INFO # DEBUG --> INFO --> WARNING --> ERROR
+    logger.setLevel(log_level)
+    format_str = '%(asctime)s %(message)s'
+    formatter = logging.Formatter(format_str)
+
+    sh = logging.StreamHandler()
+    sh.setLevel(log_level)
+    sh.setFormatter(formatter)
+    logger.addHandler(sh)
+    fh = logging.FileHandler(f"{report_name}.log", mode='w')
+    fh.setLevel(log_level)
+    fh.setFormatter(formatter)     
+    logger.addHandler(fh)
+
+    return logger
+
+def spawn_parameters(
+    flattened_parameters
+) -> List[Dict[str, Any]]:
+    algo_params : List[Dict[str, Any]] = []
+    for key in flattened_parameters:
+        _key = key.lower()
+        if _key in [ 'exchanges' ]:
+            continue
+
+        val = flattened_parameters[key]
+        if not algo_params:
+            assert(_key=="pypy_compat")
+            param_dict = {_key : val}
+            algo_params.append(param_dict)
+
+        else:
+            cloned_algo_params = None
+
+            for existing_algo_param in algo_params:
+                if type(val) not in [list, List]:
+                    existing_algo_param[_key] = val
+                else:
+                    if _key == 'hi_how_many_candles_values':
+                        for x in val:
+                            existing_algo_param['hi_stats_computed_over_how_many_candles'] = x[1]
+                            existing_algo_param['hi_candle_size'] = x[0]
+                            existing_algo_param['hi_how_many_candles'] = x[2]
+                    elif _key == 'hi_ma_short_vs_long_interval_values':
+                        for x in val:
+                            existing_algo_param['hi_ma_short_interval'] = x[0]
+                            existing_algo_param['hi_ma_long_interval'] = x[1]
+
+                    elif _key == 'lo_how_many_candles_values':
+                        for x in val:
+                            existing_algo_param['lo_stats_computed_over_how_many_candles'] = x[1]
+                            existing_algo_param['lo_candle_size'] = x[0]
+                            existing_algo_param['lo_how_many_candles'] = x[2]
+
+                    elif _key == 'lo_ma_short_vs_long_interval_values':
+                        for x in val:
+                            existing_algo_param['lo_ma_short_interval'] = x[0]
+                            existing_algo_param['lo_ma_long_interval'] = x[1]
+
+                    elif _key  in [ 'white_list_tickers', 'additional_trade_fields', 'cautious_dayofweek', 'allow_entry_dayofweek', 'mapped_event_codes', 'ecoevents_mapped_regions' ]:
+                        existing_algo_param[_key] = val
+
+                    else:
+                        if len(val)>1:
+                            cloned_algo_params = []
+
+                        if _key not in [ 'start_dates']:
+                            _key = _key.replace("_values","")
+                        elif _key == 'start_dates':
+                            _key = 'start_date'
+                            
+                        i = 0
+                        for x in val:
+                            
+                            if i==0:
+                                 existing_algo_param[_key] = x
+                            else:
+                                cloned_algo_param = existing_algo_param.copy()
+                                cloned_algo_param[_key] = x
+                                cloned_algo_params.append(cloned_algo_param)
+                            i+=1
+                        
+                        if cloned_algo_params:
+                            algo_params = algo_params + cloned_algo_params
+                            cloned_algo_params.clear()
+                            cloned_algo_params = None
+
+    param_id : int = 0
+    for algo_param in algo_params:
+        start_date = algo_param.pop('start_date')
+        name_exclude_start_date = ""
+        for key in algo_param:
+            name_exclude_start_date += f"{key}: {algo_param[key]}|"
+        name = "start_date: {start_date}|" + name_exclude_start_date
+        algo_param['param_id'] = param_id
+        algo_param['start_date'] = start_date
+        algo_param['name'] = name
+        algo_param['name_exclude_start_date'] = name_exclude_start_date
+
+        # Purpose is to avoid snowball effect in equity curves in long dated back tests.
+        if 'constant_order_notional' not in algo_param:
+            algo_param['constant_order_notional'] = True
+        algo_param['target_order_notional'] = None
+        if algo_param['constant_order_notional']:
+            algo_param['target_order_notional'] = algo_param['initial_cash'] * algo_param['entry_percent_initial_cash']/100
+        
+        param_id+=1
+    return algo_params
+
+def create_plot_canvas(key : str, pd_hi_candles : pd.DataFrame, pd_lo_candles : pd.DataFrame):
+        SMALL_SIZE = 7
+        DATE_FORMAT = '%Y-%m-%d %H:%M:%S'
+
+        plt.rc('figure', figsize=(25, 25))
+        plt.ion()
+        plt.rc('font', size=SMALL_SIZE)
+        plt.rc('axes', titlesize=SMALL_SIZE)
+        plt.rc('axes', labelsize=SMALL_SIZE)
+        plt.rc('xtick', labelsize=SMALL_SIZE)
+        plt.rc('ytick', labelsize=SMALL_SIZE)
+        plt.rc('legend', fontsize=SMALL_SIZE)
+        plt.rc('figure', titlesize=SMALL_SIZE)
+
+        fig, axes = plt.subplots(5, 1, gridspec_kw={'height_ratios': [3, 1, 1, 1, 1]})
+
+        date_numbers_hi = mdates.date2num(pd_hi_candles['datetime'])
+        date_numbers_lo = mdates.date2num(pd_lo_candles['datetime'])
+        major_locator = mdates.AutoDateLocator(minticks=3, maxticks=7)
+
+        for ax in axes:
+            ax.set_xticklabels([])
+            ax.minorticks_on()
+            ax.grid()
+            ax.xaxis.set_major_locator(major_locator)
+            ax.xaxis.set_major_formatter(mdates.DateFormatter(DATE_FORMAT))
+            ax.tick_params(axis="x", which='major', labelbottom=True, rotation=45)
+
+        time_series_canvas = axes[0]
+        time_series_canvas.minorticks_on()
+        time_series_canvas.grid()
+        time_series_canvas.set_ylabel(f'Close px and boillenger band {key}')
+        time_series_canvas.tick_params(axis="x", which='major', labelbottom=True, rotation=45)
+        time_series_canvas.plot(date_numbers_lo, pd_lo_candles['close'], color='darkblue', linewidth=2, label=f"close")
+        time_series_canvas.plot(date_numbers_hi, pd_hi_candles['boillenger_upper'], color='lightblue', linestyle='--', linewidth=0.5, label=f"upper boillenger (hi)")
+        time_series_canvas.plot(date_numbers_hi, pd_hi_candles['boillenger_lower'], color='lightblue', linestyle='--', linewidth=0.5, label=f"lower boillenger (hi)")
+        time_series_canvas.plot(date_numbers_lo, pd_lo_candles['boillenger_upper'], color='gray', linestyle='-', linewidth=1, label=f"upper boillenger (lo)")
+        time_series_canvas.plot(date_numbers_lo, pd_lo_candles['boillenger_lower'], color='gray', linestyle='-', linewidth=1, label=f"lower boillenger (lo)")
+        time_series_canvas.legend()
+
+        boillenger_channel_height_canvas = axes[1]
+        boillenger_channel_height_canvas.minorticks_on()
+        boillenger_channel_height_canvas.grid()
+        boillenger_channel_height_canvas.set_ylabel(f'boillenger channel height vs ATR band {key}')
+        boillenger_channel_height_canvas.tick_params(axis="x", which='major', labelbottom=True, rotation=45)
+        boillenger_channel_height_canvas.plot(date_numbers_hi, pd_hi_candles['boillenger_channel_height'], color='lightblue', linewidth=0.5, label=f"boillenger channel height (hi)")
+        boillenger_channel_height_canvas.plot(date_numbers_hi, pd_hi_candles['atr'], color='lightblue', linestyle='dashed', linewidth=0.5, label=f"ATR (hi)")
+        boillenger_channel_height_canvas.plot(date_numbers_lo, pd_lo_candles['boillenger_channel_height'], color='gray', linewidth=0.5, label=f"boillenger channel height (lo)")
+        boillenger_channel_height_canvas.plot(date_numbers_lo, pd_lo_candles['atr'], color='gray', linestyle='dashed', linewidth=0.5, label=f"ATR (lo)")
+        boillenger_channel_height_canvas.legend()
+
+        rsi_canvas = axes[2]
+        rsi_canvas.minorticks_on()
+        rsi_canvas.grid()
+        rsi_canvas.set_ylabel(f'RSI {key}')
+        rsi_canvas.tick_params(axis="x", which='major', labelbottom=True, rotation=45)
+        rsi_canvas.plot(date_numbers_hi, pd_hi_candles['rsi'], color='lightblue', linewidth=2, label=f"RSI (hi)")
+        rsi_canvas.plot(date_numbers_lo, pd_lo_candles['rsi'], color='gray', linestyle='dashed', linewidth=2, label=f"RSI (lo)")
+
+        macd_canvas_hi = axes[3]
+        macd_canvas_hi.minorticks_on()
+        macd_canvas_hi.grid()
+        macd_canvas_hi.set_ylabel(f'MACD hi {key}')
+        macd_canvas_hi.tick_params(axis="x", which='major', labelbottom=True, rotation=45)
+        macd_canvas_hi.plot(date_numbers_hi, pd_hi_candles['macd'], color='lightblue', linewidth=0.5, label=f"MACD (hi)")
+        macd_canvas_hi.plot(date_numbers_hi, pd_hi_candles['signal'], color='lightblue', linewidth=0.5, label=f"signal (hi)")
+        bar_colors = ['red' if value < 0 else 'green' for value in pd_hi_candles['macd_minus_signal']]
+        macd_canvas_hi.bar(date_numbers_hi, pd_hi_candles['macd_minus_signal'], width=0.005, color=bar_colors, label="MACD Histogram (hi)")
+
+        macd_canvas_lo = axes[4]
+        macd_canvas_lo.minorticks_on()
+        macd_canvas_lo.grid()
+        macd_canvas_lo.set_ylabel(f'MACD lo {key}')
+        macd_canvas_lo.tick_params(axis="x", which='major', labelbottom=True, rotation=45)
+        macd_canvas_lo.plot(date_numbers_lo, pd_lo_candles['macd'], color='gray', linewidth=0.5, label=f"MACD (lo)")
+        macd_canvas_lo.plot(date_numbers_lo, pd_lo_candles['signal'], color='gray', linewidth=0.5, label=f"signal (lo)")
+        bar_colors_lo = ['red' if value < 0 else 'green' for value in pd_lo_candles['macd_minus_signal']]
+        macd_canvas_lo.bar(date_numbers_lo, pd_lo_candles['macd_minus_signal'], width=0.005, color=bar_colors_lo, label="MACD Histogram (lo)")
+
+        return {
+            'plt' : plt,
+            'time_series_canvas' : time_series_canvas
+        }
+
+def plot_segments(
+        pd_candles : pd.DataFrame,
+        ts_partitions : Dict,
+        jpg_filename : str = None
+        ):
+    import matplotlib.pyplot as plt
+    import matplotlib.gridspec as gridspec
+
+    minima = ts_partitions['minima']
+    maxima = ts_partitions['maxima']
+    segments = ts_partitions['segments']
+
+    fig = plt.figure(figsize=(15, 8), facecolor='black')
+    gs = gridspec.GridSpec(1, 1, height_ratios=[1]) 
+
+    # Price Chart
+    ax0 = plt.subplot(gs[0])
+    ax0.plot(pd_candles['datetime'], pd_candles['close'], label='Close', color='dodgerblue')
+    ax0.plot(pd_candles['datetime'], pd_candles['smoothed_close'], label='Smoothed Close', color='yellow')
+    ax0.plot(pd_candles['datetime'], pd_candles['ema_close'], label='3m EMA', linestyle='--', color='orange')
+    ax0.fill_between(pd_candles['datetime'], pd_candles['close'], pd_candles['ema_close'], where=(pd_candles['close'] > pd_candles['ema_close']), interpolate=True, color='dodgerblue', alpha=0.3, label='Bull Market')
+    ax0.fill_between(pd_candles['datetime'], pd_candles['close'], pd_candles['ema_close'], where=(pd_candles['close'] <= pd_candles['ema_close']), interpolate=True, color='red', alpha=0.3, label='Bear Market')
+
+    ax0.set_title('Close vs EMA', color='white')
+    ax0.set_xlabel('Date', color='white')
+    ax0.set_ylabel('Price', color='white')
+    legend = ax0.legend()
+    legend.get_frame().set_facecolor('black')
+    legend.get_frame().set_edgecolor('white')
+    for text in legend.get_texts():
+        text.set_color('white')
+
+    # @CRITICAL close vs smoothed_close and merge_distance
+    for maxima_index in maxima:
+        ax0.plot(pd_candles['datetime'][maxima_index], pd_candles['close'][maxima_index], marker='+', markersize=8, color='yellow', label='maxima')
+    for minima_index in minima:
+        ax0.plot(pd_candles['datetime'][minima_index], pd_candles['close'][minima_index], marker='o', markersize=5, color='yellow', label='minima')
+
+    for segment in segments:
+        
+        ax0.axvline(x=pd_candles['datetime'][segment['end']], color='gray', linewidth=2, linestyle='--')
+
+        if 'maxima_idx_boillenger' in segment and segment['maxima_linregress_boillenger'] is not None:
+            '''
+            We don't need to compute y_series like this:
+                slope_maxima = segment['maxima_linregress_boillenger'].slope
+                intercept_maxima = segment['maxima_linregress_boillenger'].intercept
+                segment_maxima_dates = pd_candles['datetime'][segment['maxima_idx_boillenger']]
+                y_series = [ slope_maxima * idx + intercept_maxima for idx in segment['maxima_idx_boillenger'] ]
+            But, syntax is just for reference.
+            '''
+            x_series = [pd_candles.loc[idx]['datetime'] for idx in segment['maxima_idx_boillenger'] if idx in pd_candles.index] # x = dates
+            y_series = [segment['maxima_close_boillenger'][i] for i, idx in enumerate(segment['maxima_idx_boillenger']) if idx in pd_candles.index] # y = boillenger upper
+            ax0.plot(
+                    x_series,
+                    y_series,
+                color='green', linestyle='--', label='Maxima Linear Regression')
+        
+        if 'minima_idx_boillenger' in segment and segment['minima_linregress_boillenger'] is not None:
+            x_series = [pd_candles.loc[idx]['datetime'] for idx in segment['minima_idx_boillenger'] if idx in pd_candles.index] # x = dates
+            y_series = [segment['minima_close_boillenger'][i] for i, idx in enumerate(segment['minima_idx_boillenger']) if idx in pd_candles.index] # y = boillenger lower
+            ax0.plot(
+                    x_series,
+                    y_series,
+                color='red', linestyle='--', label='Minima Linear Regression')
+
+    ax0.set_facecolor('black')
+    
+    ax0.tick_params(axis='x', colors='white')
+    ax0.tick_params(axis='y', colors='white')
+    
+    # Show the plot
+    plt.grid(True)
+    plt.tight_layout()
+
+    if jpg_filename:
+        plt.savefig(jpg_filename, format='jpg', dpi=300)
+
+def segments_to_df(segments : List[Dict]) -> pd.DataFrame:
+    segments = [ 
+                            {
+                                'start' : segment['start'],
+                                'end' : segment['end'],
+                                'start_datetime' : segment['start_datetime'] if not type(segment['start_datetime']) is str else arrow.get(segment['start_datetime']).datetime.replace(tzinfo=None),
+                                'end_datetime' : segment['end_datetime'] if not type(segment['end_datetime']) is str else arrow.get(segment['end_datetime']).datetime.replace(tzinfo=None),
+                                'start_close' : segment['start_close'],
+                                'end_close' : segment['end_close'],
+                                'window_size_num_intervals' : segment['window_size_num_intervals'],
+                                'cur_recur_depth' : segment['cur_recur_depth'],
+                                'up_or_down' : segment['up_or_down'],
+                                'class' : segment['class'],
+                                'maxima_linregress_slope' : segment['maxima_linregress_full'].slope,
+                                'maxima_linregress_intercept' : segment['maxima_linregress_full'].intercept,
+                                'maxima_linregress_std_err' : segment['maxima_linregress_full'].stderr,
+                                'minima_linregress_slope' : segment['minima_linregress_full'].slope,
+                                'minima_linregress_intercept' : segment['minima_linregress_full'].intercept,
+                                'minima_linregress_std_err' : segment['minima_linregress_full'].stderr
+
+                            }
+                            for segment in segments ] 
+    for segment in segments:
+        segment['start_ts'] = int(segment['start_datetime'].timestamp())
+        segment['end_ts'] = int(segment['end_datetime'].timestamp())
+    pd_segments = pd.DataFrame(segments)
+    return pd_segments
+
+def generic_check_signal_thresholds(
+    signal_thresholds : List[Dict[str, Any]],
+    this_candle : Dict[str, Any],
+    adj_bps : float = 0
+) -> bool:
+    '''
+    WARNING!!! Do not put any strategy specific logic here!!!
+    Thanks.
+    '''
+    return all([
+        this_candle[signal['lhs']] > (this_candle[signal['rhs']] + adj_bps/10000)
+        if signal['op'] == '>' 
+        else this_candle[signal['lhs']] < (this_candle[signal['rhs']] + adj_bps/10000)
+        for signal in signal_thresholds
+    ])
+
+def generic_pnl_eval (
+            this_candle,
+            running_sl_percent_hard : float,
+            this_ticker_open_trades : List[Dict],
+            algo_param : Dict,
+            long_tp_indicator_name : str = None,
+            short_tp_indicator_name : str = None
+    ) -> Dict[str, float]:
+        '''
+        WARNING!!! Do not put any strategy specific logic here!!!
+        Thanks.
+        '''
+        unrealized_pnl_interval, unrealized_pnl_open, unrealized_pnl_live_optimistic, unrealized_pnl_live_pessimistic, unrealized_pnl_tp, unrealized_pnl_sl, unrealized_pnl_close_approx = 0, 0, 0, 0, 0, 0, 0
+        assert(len(set([ trade['side'] for trade in this_ticker_open_trades]))==1) # open trades should be in same direction
+        this_ticker_open_positions_side = this_ticker_open_trades[-1]['side']
+
+        lo_dayofweek = this_candle['dayofweek']
+        cautious_dayofweek : List[int] = algo_param['cautious_dayofweek']
+
+        lo_close = this_candle['close']
+        lo_open = this_candle['open']
+        lo_high = this_candle['high']
+        lo_low = this_candle['low']
+
+        # ATR, Fib618, bollengers are price levels. RSI/MFI..etc are not prices. Be careful.
+        long_tp_price = this_candle[long_tp_indicator_name] if long_tp_indicator_name else None
+        short_tp_price = this_candle[short_tp_indicator_name] if short_tp_indicator_name else None
+
+        _asymmetric_tp_bps = algo_param['asymmetric_tp_bps'] if lo_dayofweek in cautious_dayofweek else 0
+
+        for trade in this_ticker_open_trades:
+            target_price = trade['target_price'] if 'target_price' in trade else None
+            if not long_tp_indicator_name and not short_tp_indicator_name:
+                assert(target_price)
+
+            if this_ticker_open_positions_side=='buy':
+                unrealized_pnl_interval += (lo_close - trade['entry_price']) * trade['size']
+                unrealized_pnl_open += (lo_open - trade['entry_price']) * trade['size']
+                unrealized_pnl_live_optimistic += (lo_high - trade['entry_price']) * trade['size']
+                unrealized_pnl_live_pessimistic += (lo_low - trade['entry_price']) * trade['size']
+                unrealized_pnl_close_approx += (min(lo_close*(1+_asymmetric_tp_bps/10000), lo_high) - trade['entry_price']) * trade['size'] # Less accurate to use close price
+                if (
+                    long_tp_indicator_name 
+                    and not target_price # If entry trades are tagged target_price, it should take precedence over indicator
+                ):
+                    unrealized_pnl_tp += (min(long_tp_price*(1+_asymmetric_tp_bps/10000), lo_high) - trade['entry_price']) * trade['size']
+                else:
+                    if target_price:
+                        if (lo_high>target_price and lo_low<target_price):
+                            unrealized_pnl_tp += (target_price - trade['entry_price']) * trade['size']
+                        else:
+                            unrealized_pnl_tp += unrealized_pnl_close_approx # This is worst, try not to estimate pnl with close price!
+                    else:
+                        unrealized_pnl_tp += unrealized_pnl_close_approx # This is worst, try not to estimate pnl with close price!
+                unrealized_pnl_sl += -1 * (trade['entry_price'] * trade['size'] * (running_sl_percent_hard/100)) 
+
+            else:
+                unrealized_pnl_interval += (trade['entry_price'] - lo_close) * trade['size']
+                unrealized_pnl_open += (trade['entry_price'] - lo_open) * trade['size']
+                unrealized_pnl_live_optimistic += (trade['entry_price'] - lo_low) * trade['size']
+                unrealized_pnl_live_pessimistic += (trade['entry_price'] - lo_high) * trade['size']
+                unrealized_pnl_close_approx += (trade['entry_price'] - max(lo_close*(1-_asymmetric_tp_bps/10000), lo_low)) * trade['size']
+                if (
+                    short_tp_indicator_name 
+                    and not target_price # If entry trades are tagged target_price, it should take precedence over indicator
+                ):
+                    unrealized_pnl_tp += (trade['entry_price'] - max(short_tp_price*(1-_asymmetric_tp_bps/10000), lo_low)) * trade['size']
+                else:
+                    if target_price:
+                        if (lo_high>target_price and lo_low<target_price):
+                            unrealized_pnl_tp += (trade['entry_price'] - target_price) * trade['size']
+                        else:
+                            unrealized_pnl_tp += unrealized_pnl_close_approx # This is worst, try not to estimate pnl with close price!
+                    else:
+                        unrealized_pnl_tp += unrealized_pnl_close_approx # This is worst, try not to estimate pnl with close price!
+
+                unrealized_pnl_sl += -1 * (trade['entry_price'] * trade['size'] * (running_sl_percent_hard/100))
+
+        return {
+            'unrealized_pnl_interval' : unrealized_pnl_interval,
+            'unrealized_pnl_open' : unrealized_pnl_open,
+            'unrealized_pnl_live_optimistic' : unrealized_pnl_live_optimistic,
+            'unrealized_pnl_live_pessimistic' : unrealized_pnl_live_pessimistic,
+            'unrealized_pnl_tp' : unrealized_pnl_tp,
+            'unrealized_pnl_sl' : unrealized_pnl_sl
+        }
+
+def generic_tp_eval (
+        lo_row,
+        this_ticker_open_trades : List[Dict]
+) -> bool:
+    low : float = lo_row['low']
+    high : float = lo_row['high']
+    
+    for trade in this_ticker_open_trades:
+        if trade['target_price']<=high and trade['target_price']>=low:
+             return True
+    return False
+
+def generic_sort_filter_universe(
+    tickers : List[str],
+    exchange : Exchange,
+
+    # Use "i" (row index) to find current/last interval's market data or TAs from "all_exchange_candles"
+    i,
+    all_exchange_candles : Dict[str, Dict[str, Dict[str, pd.DataFrame]]],
+
+    max_num_tickers : int = 10
+) -> List[str]:
+    if not tickers:
+        return None
+    
+    sorted_filtered_tickers : List[str] = tickers.copy()
+
+    # Custom strategy specific sort logic here. Sort first before you filter!
+    sorted_filtered_tickers.sort()
+
+    # Custom filtering logic
+    if len(sorted_filtered_tickers)>max_num_tickers:
+         sorted_filtered_tickers = sorted_filtered_tickers[:max_num_tickers]
+
+    return sorted_filtered_tickers
+
+@retry(num_attempts=3)
+def fetch_price(
+            exchange,
+            normalized_symbol : str,
+            pd_reference_price_cache : pd.DataFrame,
+            timestamp_ms : int,
+            ref_timeframe : str = '1m'
+    ) -> float:
+    cached_row = pd_reference_price_cache[pd_reference_price_cache.timestamp_ms==timestamp_ms]
+    if cached_row.shape[0]>0:
+        reference_price = cached_row.iloc[-1]['price']
+    else:
+        reference_price = fetch_historical_price(
+                                        exchange=exchange, 
+                                        normalized_symbol=normalized_symbol, 
+                                        timestamp_ms=timestamp_ms, 
+                                        ref_timeframe=ref_timeframe)
+        cached_row = {
+            'exchange' : exchange,
+            'ticker' : normalized_symbol,
+            'datetime' : datetime.fromtimestamp(int(timestamp_ms/1000)),
+            'datetime_utc' : datetime.fromtimestamp(int(timestamp_ms/1000), tz=timezone.utc),
+            'timestamp_ms' : timestamp_ms,
+            'price' : reference_price
+        }
+        # pd_reference_price_cache = pd.concat([pd_reference_price_cache, pd.DataFrame([cached_row])], axis=0, ignore_index=True)
+        pd_reference_price_cache.loc[len(pd_reference_price_cache)] = cached_row
+    return reference_price
+
+def fetch_cycle_ath_atl(
+        exchange, 
+        symbol, 
+        timeframe, 
+        start_date : datetime, 
+        end_date : datetime
+    ):
+    ath = float('-inf')
+    atl = float('inf')
+    all_ohlcv = []
+
+    start_ts = int(start_date.timestamp() * 1000)
+    end_ts = int(end_date.timestamp() * 1000)
+    
+    while start_ts < end_ts:
+        try:
+            ohlcv = exchange.fetch_ohlcv(symbol, timeframe, since=start_ts, limit=100)
+            if not ohlcv:
+                break
+            all_ohlcv.extend(ohlcv)
+            start_ts = ohlcv[-1][0] + 1
+            time.sleep(0.1)
+        except Exception as e:
+            print(f"fetch_cycle_ath_atl Oops: {e}")
+    
+    for candle in all_ohlcv:
+        high = candle[2]
+        low = candle[3]
+        ath = max(ath, high)
+        atl = min(atl, low)
+    
+    return {
+        'ath' : ath,
+        'atl' : atl
+    }
+
+'''
+******** THE_LOOP ********
+
+This is the loop which replay candles to back tests. No STRATEGY_SPECIFIC logic should be here!!!
+'''
+def run_scenario(
+        algo_param : Dict,
+        exchanges : List[Exchange],
+        all_exchange_candles : Dict[str, Dict[str, Dict[str, pd.DataFrame]]],
+        pd_ref_candles_fast : pd.DataFrame,
+        pd_ref_candles_slow : pd.DataFrame,
+        tickers : List[str],
+        ref_candles_partitions : Dict,
+        pd_hi_candles_partitions : pd.DataFrame,
+        pd_lo_candles_partitions : pd.DataFrame,
+        economic_calendars_loaded : bool,
+        pd_economic_calendars : pd.DataFrame,
+
+        order_notional_adj_func : Callable[..., float],
+        allow_entry_initial_func : Callable[..., bool],
+        allow_entry_final_func : Callable[..., bool],
+        allow_slice_entry_func : Callable[..., bool],
+        sl_adj_func : Callable[..., Dict[str, float]],
+        trailing_stop_threshold_eval_func : Callable[..., Dict[str, float]],
+        pnl_eval_func : Callable[..., Dict[str, float]],
+        tp_eval_func : Callable[..., bool],
+        sort_filter_universe_func : Callable[..., List[str]],
+
+        logger,
+
+        pypy_compat : bool = False,
+        plot_timeseries : bool = True,
+    ):
+    exceptions : Dict = {}
+
+    if not pypy_compat:
+        pd_ref_candles_segments = segments_to_df(ref_candles_partitions['segments'])
+        pd_hi_candles_segments = segments_to_df(pd_hi_candles_partitions['segments'])
+        pd_lo_candles_segments = segments_to_df(pd_lo_candles_partitions['segments'])
+
+    min_sl_age_ms : int = 0
+    if algo_param['lo_candle_size'][-1]=="m":
+        one_interval_ms = 60*1000
+        min_sl_age_ms = algo_param['sl_num_intervals_delay'] * one_interval_ms
+        num_intervals_block_pending_ecoevents_ms = one_interval_ms*algo_param['num_intervals_block_pending_ecoevents']
+    elif algo_param['lo_candle_size'][-1]=="h":
+        one_interval_ms = 60*60*1000
+        min_sl_age_ms = algo_param['sl_num_intervals_delay'] * one_interval_ms
+        num_intervals_block_pending_ecoevents_ms = one_interval_ms*algo_param['num_intervals_block_pending_ecoevents']
+    elif algo_param['lo_candle_size'][-1]=="d":
+        one_interval_ms = 60*60*24*1000
+        min_sl_age_ms = algo_param['sl_num_intervals_delay'] * one_interval_ms
+        num_intervals_block_pending_ecoevents_ms = one_interval_ms*algo_param['num_intervals_block_pending_ecoevents']
+
+    commission_bps = algo_param['commission_bps']
+
+    initial_cash : float = algo_param['initial_cash']
+    entry_percent_initial_cash : float = algo_param['entry_percent_initial_cash']
+    target_position_size_percent_total_equity : float = algo_param['target_position_size_percent_total_equity']
+
+    class GlobalState:
+        def __init__(self, initial_cash) -> None:
+            self.cash = initial_cash
+            self.total_equity = self.cash
+            self.total_commission = 0
+
+    gloabl_state = GlobalState(initial_cash=initial_cash) # This cash position is shared across all tickers in universe
+    current_position_usdt = 0
+    
+    all_trades : List = []
+
+    compiled_candles_by_exchange_pairs : List[Dict[str, pd.DataFrame]]= {}
+    hi_num_intervals, lo_num_intervals  = 99999999, 99999999
+
+    for exchange in exchanges:
+        for ticker in tickers:
+            key : str = f"{exchange.name}-{ticker}"
+            pd_hi_candles : pd.DataFrame  = all_exchange_candles[exchange.name][ticker]['hi_candles']
+            pd_lo_candles : pd.DataFrame = all_exchange_candles[exchange.name][ticker]['lo_candles']
+
+            # market_data_gizmo sometimes insert dummy row(s) between start_date and actual first candle fetched
+            if pd_hi_candles[~pd_hi_candles.close.notna()].shape[0]>0:
+                pd_hi_candles.drop(pd_hi_candles[~pd_hi_candles.close.notna()].index[0], inplace=True)
+
+            hi_num_intervals = min(hi_num_intervals, pd_hi_candles.shape[0])
+            lo_num_intervals = min(lo_num_intervals, pd_lo_candles.shape[0])
+            
+            compiled_candles_by_exchange_pairs[key] = {}
+            compiled_candles_by_exchange_pairs[key]['hi_candles'] = pd_hi_candles
+            compiled_candles_by_exchange_pairs[key]['lo_candles'] = pd_lo_candles
+
+    all_canvas = {}
+    if plot_timeseries:
+        for exchange in exchanges:
+            for ticker in tickers:
+                key = f"{exchange.name}-{ticker}"
+                pd_hi_candles = compiled_candles_by_exchange_pairs[key]['hi_candles']
+                pd_lo_candles = compiled_candles_by_exchange_pairs[key]['lo_candles']
+
+                canvas = create_plot_canvas(key, pd_hi_candles, pd_lo_candles)
+                all_canvas[f"{key}-param_id{algo_param['param_id']}"] = canvas
+
+    order_notional_adj_func_sig = inspect.signature(order_notional_adj_func)
+    order_notional_adj_func_params = order_notional_adj_func_sig.parameters.keys()
+    allow_entry_initial_func_sig = inspect.signature(allow_entry_initial_func)
+    allow_entry_initial_func_params = allow_entry_initial_func_sig.parameters.keys()
+    allow_entry_final_func_sig = inspect.signature(allow_entry_final_func)
+    allow_entry_final_func_params = allow_entry_final_func_sig.parameters.keys()
+    allow_slice_entry_func_sig = inspect.signature(allow_slice_entry_func)
+    allow_slice_entry_func_params = allow_slice_entry_func_sig.parameters.keys()
+    sl_adj_func_sig = inspect.signature(sl_adj_func)
+    sl_adj_func_params = sl_adj_func_sig.parameters.keys()
+    trailing_stop_threshold_eval_func_sig = inspect.signature(trailing_stop_threshold_eval_func)
+    trailing_stop_threshold_eval_func_params = trailing_stop_threshold_eval_func_sig.parameters.keys()
+    tp_eval_func_sig = inspect.signature(tp_eval_func)
+    tp_eval_func_params = tp_eval_func_sig.parameters.keys()
+    sort_filter_universe_func_sig = inspect.signature(sort_filter_universe_func)
+    sort_filter_universe_func_params = sort_filter_universe_func_sig.parameters.keys()
+    
+    BUCKETS_m100_100 = bucket_series(
+						values=list([i for i in range(-100,100)]), 
+						outlier_threshold_percent=10, 
+						level_granularity=algo_param['default_level_granularity'] if 'default_level_granularity' in algo_param else 0.01
+					)
+    
+    REFERENCE_PRICE_CACHE_COLUMNS = [ 
+            'exchange', 'ticker', 'datetime', 'datetime_utc', 'timestamp_ms', 'price'
+    ]
+    reference_price_cache = {}
+
+    def _max_camp(
+            camp1 : bool,
+            camp2 : bool,
+            camp3 : bool
+    ) -> int:
+        camp : int = 1 if camp1 else 0
+        if camp2:
+            camp = 2
+        if camp3:
+            camp =3
+        return camp
+    REVERSAL_CAMP_ITEM = {
+        'camp1' : False,
+        'camp2' : False,
+        'camp3' : False,
+        'camp1_price' : None,
+        'camp2_price' : None,
+        'camp3_price' : None,
+
+        'datetime' : None # Last update
+    }
+    reversal_camp_cache = {}
+    lo_boillenger_lower_breached_cache = {}
+    lo_boillenger_upper_breached_cache = {}
+    ath, atl = None, None
+    target_order_notional = 0
+    for i in range(algo_param['how_many_last_candles'], lo_num_intervals):
+        for exchange in exchanges:
+            
+            kwargs = {k: v for k, v in locals().items() if k in sort_filter_universe_func_params}
+            sorted_filtered_tickers = sort_filter_universe_func(**kwargs)
+
+            for ticker in sorted_filtered_tickers:
+                key = f"{exchange.name}-{ticker}"
+                if key not in reversal_camp_cache:
+                    reversal_camp_cache[key] = REVERSAL_CAMP_ITEM.copy()
+
+                pd_reference_price_cache : pd.DataFrame = None
+                reference_price_cache_file : str = f"refpx_{ticker.replace('/','').replace(':','')}.csv"
+                if reference_price_cache_file not in reference_price_cache:
+                    if os.path.isfile(reference_price_cache_file):
+                        pd_reference_price_cache = pd.read_csv(reference_price_cache_file)
+                        pd_reference_price_cache.drop(pd_reference_price_cache.columns[pd_reference_price_cache.columns.str.contains('unnamed',case = False)],axis = 1, inplace = True)
+                        reference_price_cache[reference_price_cache_file] = pd_reference_price_cache
+                else:
+                    pd_reference_price_cache = reference_price_cache[reference_price_cache_file]
+                if reference_price_cache_file not in reference_price_cache:
+                    pd_reference_price_cache = pd.DataFrame(columns=REFERENCE_PRICE_CACHE_COLUMNS)
+                reference_price_cache[reference_price_cache_file] = pd_reference_price_cache
+
+                pd_candles = compiled_candles_by_exchange_pairs[key]
+                pd_hi_candles = pd_candles['hi_candles']
+                pd_lo_candles = pd_candles['lo_candles']
+                
+                lo_row = pd_lo_candles.iloc[i]
+                lo_row_tm1 = pd_lo_candles.iloc[i-1]
+
+                lo_datetime = lo_row['datetime']
+                tm1 = lo_row_tm1['datetime']
+                
+                lo_year = lo_row['year']
+                lo_month = lo_row['month']
+                lo_day = lo_row['day']
+                lo_hour = lo_row['hour']
+                lo_minute = lo_row['minute']
+                lo_timestamp_ms = lo_row['timestamp_ms']
+                lo_dayofweek = lo_row['dayofweek']
+                lo_open = lo_row['open']
+                lo_high = lo_row['high']
+                lo_low = lo_row['low']
+                lo_mid = (lo_high + lo_low)/2
+                lo_close = lo_row['close']
+                lo_candle_open_close = lo_open - lo_close
+                lo_candle_hi_lo = lo_high - lo_low
+                lo_volume = lo_row['volume']
+                lo_atr = lo_row['atr']
+                lo_rsi = lo_row['rsi']
+                lo_rsi_bucket = lo_row['rsi_bucket']
+                lo_rsi_trend = lo_row['rsi_trend']
+                lo_mfi = lo_row['mfi']
+                lo_mfi_bucket = lo_row['mfi_bucket']
+                lo_macd_minus_signal = lo_row['macd_minus_signal']
+                lo_boillenger_upper = lo_row['boillenger_upper']
+                lo_boillenger_lower = lo_row['boillenger_lower']
+                lo_boillenger_mid = (lo_boillenger_upper + lo_boillenger_lower) / 2
+                lo_boillenger_height = lo_boillenger_upper - lo_boillenger_lower 
+                lo_boillenger_channel_height = lo_row['boillenger_channel_height']
+                lo_aggressive_up = lo_row['aggressive_up']
+                lo_aggressive_down = lo_row['aggressive_down']
+                lo_fvg_high = lo_row['fvg_high']
+                lo_fvg_low = lo_row['fvg_low']
+                lo_hurst_exp = lo_row['hurst_exp']
+                lo_ema_volume_short_periods = lo_row['ema_volume_short_periods']
+                lo_ema_short_slope = lo_row['ema_short_slope'] if 'ema_short_slope' in pd_lo_candles.columns else 0
+                lo_normalized_ema_short_slope = lo_row['normalized_ema_short_slope'] if 'normalized_ema_short_slope' in pd_lo_candles.columns else 0
+                lo_ema_long_slope = lo_row['ema_long_slope'] if 'ema_long_slope' in pd_lo_candles.columns else 0
+                lo_normalized_ema_long_slope = lo_row['normalized_ema_long_slope'] if 'normalized_ema_long_slope' in pd_lo_candles.columns else 0
+                lo_tm1_normalized_ema_long_slope = lo_row_tm1['normalized_ema_long_slope'] if 'normalized_ema_long_slope' in pd_lo_candles.columns else 0
+
+                lo_tm1_close = lo_row_tm1['close']
+                lo_tm1_rsi = lo_row_tm1['rsi']
+                lo_tm1_rsi_bucket = lo_row_tm1['rsi_bucket']
+                lo_tm1_rsi_trend = lo_row_tm1['rsi_trend']
+
+                lo_max_short_periods = lo_row['max_short_periods']
+                lo_idmax_short_periods = int(lo_row['idmax_short_periods']) if not math.isnan(lo_row['idmax_short_periods']) else None
+                lo_idmax_dt_short_periods = pd_lo_candles.at[lo_idmax_short_periods, 'datetime'] if not (lo_idmax_short_periods is None or pd.isna(lo_idmax_short_periods)) else None
+                lo_max_long_periods = lo_row['max_long_periods']
+                lo_idmax_long_periods = int(lo_row['idmax_long_periods']) if not math.isnan(lo_row['idmax_long_periods']) else None
+                lo_idmax_dt_long_periods = pd_lo_candles.at[lo_idmax_long_periods, 'datetime'] if not (lo_idmax_long_periods is None or pd.isna(lo_idmax_long_periods)) else None
+
+                lo_tm1_max_short_periods = lo_row_tm1['max_short_periods']
+                lo_tm1_idmax_short_periods = int(lo_row_tm1['idmax_short_periods']) if not math.isnan(lo_row_tm1['idmax_short_periods']) else None
+                lo_tm1_idmax_dt_short_periods = pd_lo_candles.at[lo_tm1_idmax_short_periods, 'datetime'] if not (lo_tm1_idmax_short_periods is None or pd.isna(lo_tm1_idmax_short_periods)) else None
+                lo_tm1_max_long_periods = lo_row_tm1['max_long_periods']
+                lo_tm1_idmax_long_periods = int(lo_row_tm1['idmax_long_periods']) if not math.isnan(lo_row_tm1['idmax_long_periods']) else None
+                lo_tm1_idmax_dt_long_periods = pd_lo_candles.at[lo_tm1_idmax_long_periods, 'datetime'] if not (lo_tm1_idmax_long_periods is None or pd.isna(lo_tm1_idmax_long_periods)) else None
+
+                lo_min_short_periods = lo_row['min_short_periods']
+                lo_idmin_short_periods = int(lo_row['idmin_short_periods']) if not math.isnan(lo_row['idmin_short_periods']) else None
+                lo_idmin_dt_short_periods = pd_lo_candles.at[lo_idmin_short_periods,'datetime'] if not (lo_idmin_short_periods is None or pd.isna(lo_idmin_short_periods)) else None
+                lo_min_long_periods = lo_row['min_long_periods']
+                lo_idmin_long_periods = int(lo_row['idmin_long_periods']) if not math.isnan(lo_row['idmin_long_periods']) else None
+                lo_idmin_dt_long_periods = pd_lo_candles.at[lo_idmin_long_periods,'datetime'] if not (lo_idmin_long_periods is None or pd.isna(lo_idmin_long_periods)) else None
+                
+                lo_tm1_min_short_periods = lo_row_tm1['min_short_periods']
+                lo_tm1_idmin_short_periods = int(lo_row_tm1['idmin_short_periods']) if not math.isnan(lo_row_tm1['idmin_short_periods']) else None
+                lo_tm1_idmin_dt_short_periods = pd_lo_candles.at[lo_tm1_idmin_short_periods,'datetime'] if not (lo_tm1_idmin_short_periods is None or pd.isna(lo_tm1_idmin_short_periods)) else None
+                lo_tm1_min_long_periods = lo_row_tm1['min_long_periods']
+                lo_tm1_idmin_long_periods = int(lo_row_tm1['idmin_long_periods']) if not math.isnan(lo_row_tm1['idmin_long_periods']) else None
+                lo_tm1_idmin_dt_long_periods = pd_lo_candles.at[lo_tm1_idmin_long_periods,'datetime'] if not (lo_tm1_idmin_long_periods is None or pd.isna(lo_tm1_idmin_long_periods)) else None
+
+                if not ath or not atl:
+                    ath_atl = fetch_cycle_ath_atl(exchange=exchange, symbol=ticker, timeframe='1d', start_date=(algo_param['start_date'] - timedelta(days=365*4)), end_date=algo_param['start_date'])
+                    ath = ath_atl['ath']
+                    atl = ath_atl['atl']
+
+                if lo_close>ath:
+                    ath = lo_close
+                if lo_close<atl:
+                    atl = lo_close
+
+                # Incoming economic calendars? num_incoming_economic_calendars is used to Block entries if incoming events (total_num_ecoevents==0 to make entries).
+                num_impacting_economic_calendars : int = 0
+                num_bullish_ecoevents, num_bearish_ecoevents, total_num_ecoevents = 0, 0, 0
+                if economic_calendars_loaded and algo_param['block_entries_on_impacting_ecoevents']:
+                    pd_impacting_economic_calendars = pd_economic_calendars[pd_economic_calendars.event_code.isin(algo_param['mapped_event_codes'])]
+                    pd_impacting_economic_calendars = pd_impacting_economic_calendars[
+                                ( 
+                                    (
+                                        pd_impacting_economic_calendars.calendar_item_timestamp_ms>=lo_timestamp_ms) # Incoming
+                                        & (lo_timestamp_ms>=(pd_impacting_economic_calendars.calendar_item_timestamp_ms - num_intervals_block_pending_ecoevents_ms)
+                                    ) 
+                                )
+                                |
+                                ( 
+                                    (
+                                        pd_impacting_economic_calendars.calendar_item_timestamp_ms<lo_timestamp_ms) # Passed
+                                        & (lo_timestamp_ms<=(pd_impacting_economic_calendars.calendar_item_timestamp_ms + num_intervals_block_pending_ecoevents_ms/3)
+                                    ) 
+                                )
+                            ]
+                    num_impacting_economic_calendars = pd_impacting_economic_calendars.shape[0]
+
+                    if num_impacting_economic_calendars>0:
+                        pd_passed_economic_calendars = pd_impacting_economic_calendars[pd_impacting_economic_calendars.calendar_item_timestamp_ms>(lo_timestamp_ms+one_interval_ms)] # Careful with look ahead bias
+                        num_bullish_ecoevents = pd_passed_economic_calendars[pd_passed_economic_calendars.pos_neg=='bullish'].shape[0]
+                        num_bearish_ecoevents = pd_passed_economic_calendars[pd_passed_economic_calendars.pos_neg=='bearish'].shape[0]
+                        num_neutral_ecoevents = pd_passed_economic_calendars[pd_passed_economic_calendars.pos_neg=='neutral'].shape[0]
+
+                        # If adj_sl_on_ecoevents==True, total_num_ecoevents is used to set sl_percent_adj
+                        total_num_ecoevents = num_bullish_ecoevents + num_bearish_ecoevents + num_neutral_ecoevents
+                    
+                lo_fib_eval_result = lookup_fib_target(lo_row_tm1, pd_lo_candles)
+                lo_fib_short_periods_fib_target, lo_fib_short_periods_price_swing, lo_fib_long_periods_fib_target, lo_fib_long_periods_price_swing = None, None, None, None
+                if lo_fib_eval_result:
+                    lo_fib_short_periods_fib_target = lo_fib_eval_result['short_periods']['fib_target']
+                    lo_fib_long_periods_fib_target = lo_fib_eval_result['long_periods']['fib_target']
+
+                current_ref_candles_segment_index, last_ref_candles_segmment_index = -1, -1
+                current_ref_candles_segment, last_ref_candles_segment = None, None
+                current_ref_candles_segment_class, last_ref_candles_segment_class = None, None
+                if not pypy_compat:
+                    if pd_ref_candles_segments[(pd_ref_candles_segments.start_ts<=lo_datetime.timestamp()) & (pd_ref_candles_segments.end_ts>lo_datetime.timestamp()) ].shape[0]>0:
+                        current_ref_candles_segment_index = pd_ref_candles_segments[(pd_ref_candles_segments.start_ts<=lo_datetime.timestamp()) & (pd_ref_candles_segments.end_ts>lo_datetime.timestamp()) ].index.to_list()[0] # Take first
+                        current_ref_candles_segment = pd_ref_candles_segments.iloc[current_ref_candles_segment_index]
+                        if current_ref_candles_segment is not None and not current_ref_candles_segment.empty:
+                            current_ref_candles_segment_class = current_ref_candles_segment['class']
+                        last_ref_candles_segmment_index = current_ref_candles_segment_index
+                        last_ref_candles_segment = current_ref_candles_segment
+                        if current_ref_candles_segment_index>0:
+                            last_ref_candles_segmment_index = current_ref_candles_segment_index-1
+                            last_ref_candles_segment = pd_ref_candles_segments.iloc[current_ref_candles_segment_index]
+                            if last_ref_candles_segment is not None and not last_ref_candles_segment.empty:
+                                last_ref_candles_segment_class = last_ref_candles_segment['class']
+            
+                current_hi_candles_segment_index, last_hi_candles_segmment_index = -1, -1
+                current_hi_candles_segment, last_hi_candles_segment = None, None
+                current_hi_candles_segment_class, last_hi_candles_segment_class = None, None
+                if not pypy_compat:
+                    if pd_hi_candles_segments[(pd_hi_candles_segments.start_ts<=lo_datetime.timestamp()) & (pd_hi_candles_segments.end_ts>lo_datetime.timestamp()) ].shape[0]>0:
+                        current_hi_candles_segment_index = pd_hi_candles_segments[(pd_hi_candles_segments.start_ts<=lo_datetime.timestamp()) & (pd_hi_candles_segments.end_ts>lo_datetime.timestamp()) ].index.to_list()[0] # Take first
+                        current_hi_candles_segment = pd_hi_candles_segments.iloc[current_hi_candles_segment_index]
+                        if current_hi_candles_segment is not None and not current_hi_candles_segment.empty:
+                            current_hi_candles_segment_class = current_hi_candles_segment['class']
+                        last_hi_candles_segmment_index = current_hi_candles_segment_index
+                        last_hi_candles_segment = current_hi_candles_segment
+                        if current_hi_candles_segment_index>0:
+                            last_hi_candles_segmment_index = current_hi_candles_segment_index-1
+                            last_hi_candles_segment = pd_hi_candles_segments.iloc[current_hi_candles_segment_index]
+                            if last_hi_candles_segment is not None and not last_hi_candles_segment.empty:
+                                last_hi_candles_segment_class = last_hi_candles_segment['class']
+
+                current_lo_candles_segment_index, last_lo_candles_segmment_index = -1, -1
+                current_lo_candles_segment, last_lo_candles_segment = None, None
+                current_lo_candles_segment_class, last_lo_candles_segment_class = None, None
+                if not pypy_compat:
+                    if pd_lo_candles_segments[(pd_lo_candles_segments.start_ts<=lo_datetime.timestamp()) & (pd_lo_candles_segments.end_ts>lo_datetime.timestamp()) ].shape[0]>0:
+                        current_lo_candles_segment_index = pd_lo_candles_segments[(pd_lo_candles_segments.start_ts<=lo_datetime.timestamp()) & (pd_lo_candles_segments.end_ts>lo_datetime.timestamp()) ].index.to_list()[0] # Take first
+                        current_lo_candles_segment = pd_lo_candles_segments.iloc[current_lo_candles_segment_index]
+                        if current_lo_candles_segment is not None and not current_lo_candles_segment.empty:
+                            current_lo_candles_segment_class = current_lo_candles_segment['class']
+                        last_lo_candles_segmment_index = current_lo_candles_segment_index
+                        last_lo_candles_segment = current_lo_candles_segment
+                        if current_lo_candles_segment_index>0:
+                            last_lo_candles_segmment_index = current_lo_candles_segment_index-1
+                            last_lo_candles_segment = pd_lo_candles_segments.iloc[current_lo_candles_segment_index]
+                            if last_lo_candles_segment is not None and not last_lo_candles_segment.empty: 
+                                last_lo_candles_segment_class = last_lo_candles_segment['class']
+
+                # Find corresponding row in pd_hi_candles
+                def _find_ref_row(lo_year, lo_month, lo_day, pd_ref_candles):
+                    ref_row = None
+                    ref_matching_rows = pd_ref_candles[(pd_ref_candles.year==lo_year) & (pd_ref_candles.month==lo_month) & (pd_ref_candles.day==lo_day)]
+                    if not ref_matching_rows.empty:
+                        ref_row = ref_matching_rows.iloc[0]
+                        ref_row.has_inflection_point = False
+
+                        recent_rows = pd_ref_candles[(pd_ref_candles['datetime'] <= lo_datetime)].tail(3)
+                        if not recent_rows['close_above_or_below_ema'].isna().all():
+                            ref_row.has_inflection_point = True
+
+                    else:
+                        logger.warning(f"{key} ref_row not found for year: {lo_year}, month: {lo_month}, day: {lo_day}")
+
+                    return ref_row
+                
+                def _search_hi_tm1(hi_row, lo_row, pd_hi_candles):
+                    row_index = hi_row.name -1
+                    hi_row_tm1 = pd_hi_candles.iloc[row_index] if hi_row is not None else None
+                    hi_row_tm1 = hi_row_tm1 if hi_row_tm1['timestamp_ms'] < lo_row['timestamp_ms'] else None
+                    if row_index>1:
+                        while hi_row_tm1['timestamp_ms'] >= lo_row['timestamp_ms']:
+                            row_index = row_index -1
+                            hi_row_tm1 = pd_hi_candles.iloc[row_index]
+                    return hi_row_tm1
+
+                hi_row, hi_row_tm1 = None, None
+                if lo_datetime>=algo_param['start_date']:
+                    if algo_param['lo_candle_size'][-1]=="m":
+                        matching_rows = pd_hi_candles[(pd_hi_candles.year==lo_year) & (pd_hi_candles.month==lo_month) & (pd_hi_candles.day==lo_day) & (pd_hi_candles.hour==lo_hour)]
+                        if not matching_rows.empty:
+                            hi_row = matching_rows.iloc[0]
+                            
+                        else:
+                            logger.warning(f"{key} hi_row not found for year: {lo_year}, month: {lo_month}, day: {lo_day}, hour: {lo_hour}")
+                            continue
+
+                        hi_row_tm1 = _search_hi_tm1(hi_row, lo_row, pd_hi_candles)
+                        if hi_row_tm1 is not None:
+                            assert(hi_row_tm1['timestamp_ms'] < lo_row['timestamp_ms']) # No look ahead bias!!!
+                        else:
+                            continue
+
+                        # Be careful with look ahead bias!!!
+                        target_ref_candle_date = lo_datetime + timedelta(days=-1)
+                        ref_row_fast = _find_ref_row(
+                            target_ref_candle_date.year, 
+                            target_ref_candle_date.month, 
+                            target_ref_candle_date.day, 
+                            pd_ref_candles_fast)
+                        ref_row_slow = _find_ref_row(
+                            target_ref_candle_date.year, 
+                            target_ref_candle_date.month, 
+                            target_ref_candle_date.day, 
+                            pd_ref_candles_slow)
+
+                    elif algo_param['lo_candle_size'][-1]=="h":
+                        matching_rows = pd_hi_candles[(pd_hi_candles.year==lo_year) & (pd_hi_candles.month==lo_month) & (pd_hi_candles.day==lo_day)]
+                        if not matching_rows.empty:
+                            hi_row = matching_rows.iloc[0]
+                            
+                        else:
+                            logger.warning(f"{key} hi_row not found for year: {lo_year}, month: {lo_month}, day: {lo_day}")
+                            continue
+
+                        hi_row_tm1 = _search_hi_tm1(hi_row, lo_row, pd_hi_candles)
+                        if hi_row_tm1 is not None:
+                            assert(hi_row_tm1['timestamp_ms'] < lo_row['timestamp_ms']) # No look ahead bias!!!
+                        else:
+                            continue
+                        
+                        # Be careful with look ahead bias!!!
+                        target_ref_candle_date = lo_datetime + timedelta(days=-1)
+                        ref_row_fast = _find_ref_row(
+                            target_ref_candle_date.year, 
+                            target_ref_candle_date.month, 
+                            target_ref_candle_date.day, 
+                            pd_ref_candles_fast)
+                        ref_row_slow = _find_ref_row(
+                            target_ref_candle_date.year, 
+                            target_ref_candle_date.month, 
+                            target_ref_candle_date.day, 
+                            pd_ref_candles_slow)
+
+                    elif algo_param['lo_candle_size'][-1]=="d":
+                        # Not supported atm
+                        hi_row, hi_row_tm1 = None, None
+
+                    hi_datetime, hi_year, hi_month, hi_day, hi_hour, hi_minute, hi_timestamp_ms = None, None, None, None, None, None, None
+                    hi_open, hi_high, hi_low, hi_close, hi_volume = None, None, None, None, None
+                    hi_atr, hi_rsi, hi_rsi_bucket, hi_rsi_trend, hi_mfi, hi_mfi_bucket, hi_macd_minus_signal, hi_boillenger_upper, hi_boillenger_lower, hi_boillenger_channel_height = None, None, None, None, None, None, None, None, None, None
+                    hi_hurst_exp, hi_ema_volume_long_periods = None, None
+                    hi_tm1_rsi, hi_tm1_rsi_bucket, hi_tm1_rsi_trend = None, None, None
+                    hi_fib_eval_result = None
+                    if hi_row is not None:
+                        hi_datetime = hi_row['datetime']
+                        hi_year = hi_row['year']
+                        hi_month = hi_row['month']
+                        hi_day = hi_row['day']
+                        hi_hour = hi_row['hour']
+                        hi_minute = hi_row['minute']
+                        hi_timestamp_ms = hi_row['timestamp_ms']
+                        hi_open = hi_row['open']
+                        hi_high = hi_row['high']
+                        hi_low = hi_row['low']
+                        hi_close = hi_row['close']
+                        hi_volume = hi_row['volume']
+                        hi_atr = hi_row['atr']
+                        hi_rsi = hi_row['rsi']
+                        hi_rsi_bucket = hi_row['rsi_bucket']
+                        hi_rsi_trend = hi_row['rsi_trend']
+                        hi_mfi = hi_row['mfi']
+                        hi_mfi_bucket = hi_row['mfi_bucket']
+                        hi_macd_minus_signal = hi_row['macd_minus_signal']
+                        hi_boillenger_upper = hi_row['boillenger_upper']
+                        hi_boillenger_lower = hi_row['boillenger_lower']
+                        hi_boillenger_channel_height = hi_row['boillenger_channel_height']
+                        hi_hurst_exp = hi_row['hurst_exp']
+
+                        hi_tm1_rsi = hi_row_tm1['rsi']
+                        hi_tm1_rsi_bucket = hi_row_tm1['rsi_bucket']
+                        hi_tm1_rsi_trend = hi_row_tm1['rsi_trend']
+
+                        hi_ema_volume_long_periods = hi_row['ema_volume_long_periods']
+                        hi_ema_short_slope = hi_row['ema_short_slope'] if 'ema_short_slope' in pd_hi_candles.columns else 0
+                        hi_normalized_ema_short_slope = hi_row['normalized_ema_short_slope'] if 'normalized_ema_short_slope' in pd_hi_candles.columns else 0
+                        hi_ema_long_slope = hi_row['ema_long_slope'] if 'ema_long_slope' in pd_hi_candles.columns else 0
+                        hi_normalized_ema_long_slope = hi_row['normalized_ema_long_slope'] if 'normalized_ema_long_slope' in pd_hi_candles.columns else 0
+                        hi_tm1_normalized_ema_long_slope = hi_row_tm1['normalized_ema_long_slope'] if 'normalized_ema_long_slope' in pd_hi_candles.columns else 0
+
+                        hi_max_short_periods = hi_row['max_short_periods']
+                        hi_idmax_short_periods = int(hi_row['idmax_short_periods']) if not math.isnan(hi_row['idmax_short_periods']) else None
+                        hi_idmax_dt_short_periods = pd_hi_candles.at[hi_idmax_short_periods,'datetime'] if not(hi_idmax_short_periods is None or pd.isna(hi_idmax_short_periods)) else None
+                        hi_max_long_periods = hi_row['max_long_periods']
+                        hi_idmax_long_periods = int(hi_row['idmax_long_periods']) if not math.isnan(hi_row['idmax_long_periods']) else None
+                        hi_idmax_dt_long_periods = pd_hi_candles.at[hi_idmax_long_periods,'datetime'] if not(hi_idmax_long_periods is None or pd.isna(hi_idmax_long_periods)) else None
+
+                        hi_tm1_max_short_periods = hi_row_tm1['max_short_periods']
+                        hi_tm1_idmax_short_periods = int(hi_row_tm1['idmax_short_periods']) if not math.isnan(hi_row_tm1['idmax_short_periods']) else None
+                        hi_tm1_idmax_dt_short_periods = pd_hi_candles.at[hi_tm1_idmax_short_periods,'datetime'] if not(hi_tm1_idmax_short_periods is None or pd.isna(hi_tm1_idmax_short_periods)) else None
+                        hi_tm1_max_long_periods = hi_row_tm1['max_long_periods']
+                        hi_tm1_idmax_long_periods = int(hi_row_tm1['idmax_long_periods']) if not math.isnan(hi_row_tm1['idmax_long_periods']) else None
+                        hi_tm1_idmax_dt_long_periods = pd_hi_candles.at[hi_tm1_idmax_long_periods,'datetime'] if not(hi_tm1_idmax_long_periods is None or pd.isna(hi_tm1_idmax_long_periods)) else None
+
+                        hi_min_short_periods = hi_row['min_short_periods']
+                        hi_idmin_short_periods = int(hi_row['idmin_short_periods']) if not math.isnan(hi_row['idmin_short_periods']) else None
+                        hi_idmin_dt_short_periods = pd_hi_candles.at[hi_idmin_short_periods,'datetime'] if not (hi_idmin_short_periods is None or pd.isna(hi_idmin_short_periods)) else None
+                        hi_min_long_periods = hi_row['min_long_periods']
+                        hi_idmin_long_periods = int(hi_row['idmin_long_periods']) if not math.isnan(hi_row['idmin_long_periods']) else None
+                        hi_idmin_dt_long_periods = pd_hi_candles.at[hi_idmin_long_periods,'datetime'] if not (hi_idmin_long_periods is None or pd.isna(hi_idmin_long_periods)) else None
+
+                        hi_tm1_min_short_periods = hi_row_tm1['min_short_periods']
+                        hi_tm1_idmin_short_periods = int(hi_row_tm1['idmin_short_periods']) if not math.isnan(hi_row_tm1['idmin_short_periods']) else None
+                        hi_tm1_idmin_dt_short_periods = pd_hi_candles.at[hi_tm1_idmin_short_periods,'datetime'] if not (hi_tm1_idmin_short_periods is None or pd.isna(hi_tm1_idmin_short_periods)) else None
+                        hi_tm1_min_long_periods = hi_row_tm1['min_long_periods']
+                        hi_tm1_idmin_long_periods = int(hi_row_tm1['idmin_long_periods']) if not math.isnan(hi_row_tm1['idmin_long_periods']) else None
+                        hi_tm1_idmin_dt_long_periods = pd_hi_candles.at[hi_tm1_idmin_long_periods,'datetime'] if not (hi_tm1_idmin_long_periods is None or pd.isna(hi_tm1_idmin_long_periods)) else None
+
+                        hi_fib_eval_result = lookup_fib_target(hi_row_tm1, pd_hi_candles)
+                        hi_fib_short_periods_fib_target, hi_fib_short_periods_price_swing, hi_fib_long_periods_fib_target, hi_fib_long_periods_price_swing = None, None, None, None
+                        if hi_fib_eval_result:
+                            hi_fib_short_periods_fib_target = hi_fib_eval_result['short_periods']['fib_target']
+                            hi_fib_long_periods_fib_target = hi_fib_eval_result['long_periods']['fib_target']
+
+                    last_candles, post_move_candles, post_move_price_change, post_move_price_change_percent = None, None, None, None
+                    if algo_param['last_candles_timeframe']=='lo':
+                        last_candles = pd_lo_candles[pd_lo_candles['timestamp_ms']<=lo_timestamp_ms].tail(algo_param['how_many_last_candles']).to_dict('records')
+                        assert(all([ candle['timestamp_ms']<=lo_timestamp_ms for candle in last_candles ]))
+                        post_move_candles = pd_lo_candles[pd_lo_candles['timestamp_ms']<=lo_timestamp_ms].tail(algo_param['post_move_num_intervals']).to_dict('records')
+                        
+                    elif algo_param['last_candles_timeframe']=='hi' and hi_row is not None:
+                        last_candles = pd_hi_candles[pd_hi_candles['timestamp_ms']<=hi_timestamp_ms].tail(algo_param['how_many_last_candles']).to_dict('records')
+                        assert(all([ candle['timestamp_ms']<=hi_timestamp_ms for candle in last_candles ]))
+                        post_move_candles = pd_hi_candles[pd_hi_candles['timestamp_ms']<=hi_timestamp_ms].tail(algo_param['post_move_num_intervals']).to_dict('records')
+
+                    post_move_price_change, post_move_price_change_percent = 0, 0
+                    if post_move_candles and len(post_move_candles)>=2:
+                        post_move_price_change = post_move_candles[-1]['close'] - post_move_candles[0]['open']
+                        post_move_price_change_percent = 0
+                        if post_move_price_change>0:
+                            post_move_price_change_percent = (post_move_candles[-1]['close']/post_move_candles[0]['open'] -1) * 100
+                        else:
+                            post_move_price_change_percent = -(post_move_candles[0]['close']/post_move_candles[-1]['open'] -1) * 100
+
+                    ref_close_fast, ref_ema_close_fast = None, None
+                    if ref_row_fast is not None:
+                        ref_close_fast = ref_row_fast['close']
+                        ref_ema_close_fast = ref_row_fast['ema_close']
+
+                    ref_close_slow, ref_ema_close_slow = None, None
+                    if ref_row_slow is not None:
+                        ref_close_slow = ref_row_slow['close']
+                        ref_ema_close_slow = ref_row_slow['ema_close']
+
+                    # POSITION NOTIONAL MARKING lo_low, lo_high. pessimistic!
+                    def _refresh_current_position(timestamp_ms):
+                        current_position_usdt_buy = sum([x['size'] * lo_close for x in all_trades if not x['closed'] and x['side']=='buy'])
+                        current_position_usdt_sell = sum([x['size'] * lo_close for x in all_trades if not x['closed'] and x['side']=='sell'])
+                        current_position_usdt = current_position_usdt_buy + current_position_usdt_sell
+                        this_ticker_historical_trades = [ trade for trade in all_trades if trade['symbol']==ticker ]
+                        this_ticker_open_trades = [ trade for trade in this_ticker_historical_trades if not trade['closed'] ]
+                        this_ticker_current_position_usdt_buy = sum([x['size'] * lo_close for x in this_ticker_open_trades if x['side']=='buy'])
+                        this_ticker_current_position_usdt_sell = sum([x['size'] * lo_close for x in this_ticker_open_trades if x['side']=='sell'])
+
+                        entries_since_sl : Union[int, None] = -1
+
+                        avg_entry_price = None
+                        pos_side = '---'
+                        max_trade_age_ms = timestamp_ms
+                        if this_ticker_open_trades:
+                            max_trade_age_ms = timestamp_ms - max([trade['timestamp_ms'] for trade in this_ticker_open_trades ])
+
+                            avg_entry_price = sum([ trade['entry_price']*trade['size'] for trade in this_ticker_open_trades]) / sum([ trade['size'] for trade in this_ticker_open_trades])
+
+                            sides = [ x['side'] for x in this_ticker_open_trades ]
+                            if len(set(sides))==1:
+                                if sides[0]=='buy':
+                                    pos_side = 'buy'
+                                else:
+                                    pos_side = 'sell'
+
+                        max_sl_trade_age_ms = None
+                        this_ticker_sl_trades = [ trade for trade in this_ticker_historical_trades if trade['reason']=='SL' ]
+                        if this_ticker_sl_trades:
+                            last_sl_timestamp_ms = max([trade['timestamp_ms'] for trade in this_ticker_sl_trades ])
+                            max_sl_trade_age_ms = timestamp_ms - last_sl_timestamp_ms
+                            entries_since_sl = len([trade for trade in this_ticker_historical_trades if trade['timestamp_ms']>last_sl_timestamp_ms and trade['reason']=='entry' and trade['closed']] )
+
+                        # In single legged trading, we either long or short for a particular ticker at any given moment
+                        assert(
+                            (this_ticker_current_position_usdt_buy>=0 and this_ticker_current_position_usdt_sell==0) 
+                            or (this_ticker_current_position_usdt_buy==0 and this_ticker_current_position_usdt_sell>=0))
+                        
+                        if this_ticker_current_position_usdt_buy>0:
+                            this_ticker_open_positions_side = 'buy'
+                            this_ticker_current_position_usdt = this_ticker_current_position_usdt_buy
+                        elif this_ticker_current_position_usdt_sell>0:
+                            this_ticker_open_positions_side = 'sell'
+                            this_ticker_current_position_usdt = this_ticker_current_position_usdt_sell
+                        else:
+                            this_ticker_open_positions_side = 'flat'
+                            this_ticker_current_position_usdt = 0
+
+                        return {
+                            'avg_entry_price' : avg_entry_price,
+                            'side' : pos_side,
+                            'current_position_usdt_buy' : current_position_usdt_buy,
+                            'current_position_usdt_sell' : current_position_usdt_sell,
+                            'current_position_usdt' : current_position_usdt,
+                            'this_ticker_open_trades' : this_ticker_open_trades,
+                            'this_ticker_current_position_usdt_buy' : this_ticker_current_position_usdt_buy,
+                            'this_ticker_current_position_usdt_sell' : this_ticker_current_position_usdt_sell,
+                            'this_ticker_open_positions_side' : this_ticker_open_positions_side,
+                            'this_ticker_current_position_usdt' : this_ticker_current_position_usdt,
+                            'max_trade_age_ms' : max_trade_age_ms,
+                            'max_sl_trade_age_ms' : max_sl_trade_age_ms,
+                            'entries_since_sl' : entries_since_sl
+                        }
+                    
+                    current_positions_info = _refresh_current_position(lo_timestamp_ms)
+                    avg_entry_price = current_positions_info['avg_entry_price']
+                    pos_side = current_positions_info['side']
+                    current_position_usdt_buy = current_positions_info['current_position_usdt_buy']
+                    current_position_usdt_sell = current_positions_info['current_position_usdt_sell']
+                    current_position_usdt = current_positions_info['current_position_usdt']
+                    this_ticker_open_trades = current_positions_info['this_ticker_open_trades']
+                    this_ticker_current_position_usdt_buy = current_positions_info['this_ticker_current_position_usdt_buy']
+                    this_ticker_current_position_usdt_sell = current_positions_info['this_ticker_current_position_usdt_sell']
+                    this_ticker_open_positions_side = current_positions_info['this_ticker_open_positions_side']
+                    this_ticker_current_position_usdt = current_positions_info['this_ticker_current_position_usdt']
+                    max_trade_age_ms = current_positions_info['max_trade_age_ms']
+                    max_sl_trade_age_ms = current_positions_info['max_sl_trade_age_ms']
+                    entries_since_sl = current_positions_info['entries_since_sl']
+                    block_entry_since_last_sl = True if max_sl_trade_age_ms and max_sl_trade_age_ms<=min_sl_age_ms else False
+
+                    def _close_open_positions(
+                            key, ticker, 
+                            this_ticker_current_position_usdt, 
+                            this_ticker_open_positions_side, 
+                            current_position_usdt, 
+                            trade_pnl,
+                            effective_tp_trailing_percent, 
+                            row, 
+                            reason, 
+                            reason2,
+                            gloabl_state, all_trades, all_canvas,
+                            algo_param,
+                            standard_pnl_percent_buckets=BUCKETS_m100_100
+                        ):
+                        def _gains_losses_to_label(gains_losses_percent):
+                            gains_losses_percent_label = bucketize_val(gains_losses_percent, buckets=standard_pnl_percent_buckets)
+
+                            if gains_losses_percent>=0:
+                                return f"gain {gains_losses_percent_label}%"
+                            else:
+                                return f"loss {gains_losses_percent_label}%"
+                        
+                        def _how_long_before_closed_sec_to_label(how_long_before_closed_sec):
+                            how_long_before_closed_sec_label = None
+                            how_long_before_closed_hr = how_long_before_closed_sec/(60*60)
+                            if how_long_before_closed_hr<=1:
+                                how_long_before_closed_sec_label = "<=1hr"
+                            elif how_long_before_closed_hr>1 and how_long_before_closed_hr<=8:
+                                how_long_before_closed_sec_label = ">1hr <=8hr"
+                            elif how_long_before_closed_hr>8 and how_long_before_closed_hr<=24:
+                                how_long_before_closed_sec_label = ">8hr <=24hr"
+                            elif how_long_before_closed_hr>24 and how_long_before_closed_hr<=24*7:
+                                how_long_before_closed_sec_label = ">24hr <=7days"
+                            elif how_long_before_closed_hr>24*7 and how_long_before_closed_hr<=24*7*2:
+                                how_long_before_closed_sec_label = ">7days <=14days"
+                            else:
+                                how_long_before_closed_sec_label = ">14days"
+                            return how_long_before_closed_sec_label
+                        
+                        this_ticker_open_trades = [ trade for trade in all_trades if not trade['closed'] and trade['symbol']==ticker]
+
+                        entry_dt = min([ trade['trade_datetime'] for trade in this_ticker_open_trades ])
+                        entry_dayofweek = entry_dt.dayofweek
+                        entry_hour = entry_dt.hour
+
+                        this_datetime = row['datetime']
+                        this_timestamp_ms = row['timestamp_ms']
+                        dayofweek = row['dayofweek']
+                        high = row['high']
+                        low = row['low']
+                        close = row['close']
+                        ema_short_slope = row['ema_short_slope'] if 'ema_short_slope' in row else None
+                        ema_long_slope = row['ema_long_slope'] if 'ema_long_slope' in row else None
+
+                        # Step 1. mark open trades as closed first
+                        entry_commission, exit_commission = 0, 0
+                        for trade in this_ticker_open_trades:
+                            entry_commission += trade['commission']
+                            if this_ticker_open_positions_side=='buy':
+                                exit_commission += close * trade['size'] * commission_bps / 10000
+                                
+                            else:
+                                exit_commission += close * trade['size'] * commission_bps / 10000
+                            trade['trade_pnl'] = 0 # trade_pnl parked under closing trade
+                            trade['trade_pnl_bps'] = 0
+                            trade['closed'] = True
+                        max_pain = min([ trade['max_pain'] for trade in this_ticker_open_trades])
+                        max_pain_percent = max_pain/this_ticker_current_position_usdt * 100
+                        max_pain_percent_label = _gains_losses_to_label(max_pain_percent)
+
+                        timestamp_ms_from_closed_trades = min([ trade['timestamp_ms'] for trade in this_ticker_open_trades])
+                        num_impacting_economic_calendars = min([ trade['num_impacting_economic_calendars'] if 'num_impacting_economic_calendars' in trade else 0 for trade in this_ticker_open_trades])
+                        max_camp = max([ trade['max_camp'] for trade in this_ticker_open_trades])
+                        entry_post_move_price_change_percent = max([ trade['post_move_price_change_percent'] if 'post_move_price_change_percent' in trade else 0 for trade in this_ticker_open_trades ])
+                        
+                        # Step 2. Update global_state
+                        trade_pnl_less_comm = trade_pnl - (entry_commission + exit_commission)
+                        gains_losses_percent = trade_pnl_less_comm/this_ticker_current_position_usdt * 100
+                        gains_losses_percent_label = _gains_losses_to_label(gains_losses_percent)
+                        how_long_before_closed_sec = (this_timestamp_ms - timestamp_ms_from_closed_trades) / 1000
+                        how_long_before_closed_sec_label = _how_long_before_closed_sec_to_label(how_long_before_closed_sec)
+
+                        gloabl_state.total_equity += trade_pnl_less_comm
+                        gloabl_state.total_commission += exit_commission
+                        cash_before = gloabl_state.cash
+                        gloabl_state.cash = gloabl_state.total_equity
+                        cash_after = gloabl_state.cash
+                        running_total_num_positions : int = len([ 1 for x in all_trades if x['reason']=='entry' and not x['closed']])
+
+                        # Step 3. closing trade
+                        # closing_price = low if this_ticker_open_positions_side=='buy' else high # pessimistic!
+                        closing_price = close
+                        closing_trade = {
+                            'trade_datetime' : this_datetime,
+                            'timestamp_ms' : this_timestamp_ms,
+                            'dayofweek' : dayofweek,
+                            'entry_dt' : entry_dt,
+                            'entry_dayofweek' : entry_dayofweek,
+                            'entry_hour' : entry_hour,
+                            'exchange' : exchange.name,
+                            'symbol' : ticker,
+                            'side' : 'sell' if this_ticker_open_positions_side=='buy' else 'buy',
+                            'size' : this_ticker_current_position_usdt / closing_price, # in base ccy
+                            'entry_price' : closing_price, # pessimistic!
+                            'closed' : True,
+                            'reason' : reason,
+                            'reason2' : reason2,
+                            'total_equity' : gloabl_state.total_equity,
+                            'this_ticker_current_position_usdt' : this_ticker_current_position_usdt,
+                            'current_position_usdt' : current_position_usdt,
+                            'running_total_num_positions' : running_total_num_positions,
+                            'cash_before' : cash_before,
+                            'cash_after' : cash_after,
+                            'order_notional' : this_ticker_current_position_usdt,
+                            'trade_pnl' : trade_pnl,
+                            'commission' : exit_commission,
+                            'max_pain' : max_pain,
+                            'trade_pnl_less_comm': trade_pnl_less_comm,
+                            'trade_pnl_bps' : (trade_pnl / this_ticker_current_position_usdt) * 100 * 100 if this_ticker_current_position_usdt!=0 else 0,
+                            'gains_losses_percent' : gains_losses_percent,
+                            'gains_losses_percent_label' : gains_losses_percent_label,
+                            'how_long_before_closed_sec' : how_long_before_closed_sec,
+                            'how_long_before_closed_sec_label' : how_long_before_closed_sec_label,
+                            'max_pain_percent' : max_pain_percent,
+                            'max_pain_percent_label' : max_pain_percent_label,
+                            'ema_short_slope' : ema_short_slope,
+                            'ema_long_slope' : ema_long_slope,
+                            'num_impacting_economic_calendars' : num_impacting_economic_calendars,
+                            'max_camp' : max_camp,
+                            'entry_post_move_price_change_percent' : entry_post_move_price_change_percent
+                        }
+                        _last_open_trade = this_ticker_open_trades[-1]
+                        additional_fields = {field: _last_open_trade[field] if field in _last_open_trade else None for field in algo_param['additional_trade_fields']}
+                        closing_trade.update(additional_fields)
+                        all_trades.append(closing_trade)
+
+                        if plot_timeseries:
+                            '''
+                            https://matplotlib.org/stable/api/_as_gen/matplotlib.axes.Axes.axvline.html
+                            linestyle='-' means solid line. If you don't supply linestyle, the vertical line wont show!!!
+                            '''
+                            color = 'green' if reason=='TP' or (reason=='HC' and trade_pnl_less_comm>0) else 'red'    
+                            all_canvas[f"{key}-param_id{algo_param['param_id']}"]['time_series_canvas'].axvline(x=this_datetime, color=color, linewidth=2, linestyle='--')
+                            all_canvas[f"{key}-param_id{algo_param['param_id']}"]['time_series_canvas'].scatter([this_datetime, this_datetime], [low, high], color=color)
+                    
+                    # UNREAL EVALUATION. We're being pessimistic! We use low/high for estimating unrealized_pnl for buys and sells respectively here.
+                    pnl_percent_notional = 0
+                    if current_position_usdt>0:    
+                        unrealized_pnl, unrealized_pnl_interval, unrealized_pnl_open, unrealized_pnl_live_optimistic, unrealized_pnl_live_pessimistic, unrealized_pnl_live, max_pnl_percent_notional, unrealized_pnl_boillenger, unrealized_pnl_sl, max_unrealized_pnl_live, max_pain, recovered_pnl_optimistic, recovered_pnl_pessimistic, max_recovered_pnl = 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 # USDT
+                        _asymmetric_tp_bps = algo_param['asymmetric_tp_bps'] if lo_dayofweek in algo_param['cautious_dayofweek'] else 0
+                        
+                        max_unrealized_pnl_live = max([ trade['max_unrealized_pnl_live'] if 'max_unrealized_pnl_live' in trade else 0 for trade in this_ticker_open_trades ])
+                        # 'min' max_pain becaues max_pain is a negative number. It's a loss!
+                        max_pain = min([ trade['max_pain'] if 'max_pain' in trade else 0 for trade in this_ticker_open_trades ])
+                        max_recovered_pnl = max([ trade['max_recovered_pnl'] if 'max_recovered_pnl' in trade else 0 for trade in this_ticker_open_trades ])
+                        trade_datetime = max([ trade['trade_datetime'] if 'trade_datetime' in trade else 0 for trade in this_ticker_open_trades ])
+                        entry_post_move_price_change_percent = max([ trade['post_move_price_change_percent'] if 'post_move_price_change_percent' in trade else 0 for trade in this_ticker_open_trades ])
+                        max_camp = max([ trade['max_camp'] for trade in this_ticker_open_trades])
+                        running_sl_percent_hard = max([ trade['running_sl_percent_hard'] for trade in this_ticker_open_trades])
+
+                        max_pnl_potential_percent = None
+                        if any([ trade for trade in this_ticker_open_trades if 'target_price' in trade ]):
+                            max_pnl_potential_percent = max([ (trade['target_price']/trade['entry_price'] -1) *100 if trade['side']=='buy' else (trade['entry_price']/trade['target_price'] -1) *100 for trade in this_ticker_open_trades if 'target_price' in trade ])
+
+                        kwargs = {k: v for k, v in locals().items() if k in sl_adj_func_params}
+                        sl_adj_func_result = sl_adj_func(**kwargs)
+                        running_sl_percent_hard = sl_adj_func_result['running_sl_percent_hard']
+
+                        # this_ticker_open_trades should be updated after SL adj eval
+                        for trade in this_ticker_open_trades:
+                            trade['running_sl_percent_hard'] = running_sl_percent_hard
+                        
+                        kwargs = {k: v for k, v in locals().items() if k in trailing_stop_threshold_eval_func_params}
+                        trailing_stop_threshold_eval_func_result = trailing_stop_threshold_eval_func(**kwargs)
+                        tp_min_percent = trailing_stop_threshold_eval_func_result['tp_min_percent']
+                        tp_max_percent = trailing_stop_threshold_eval_func_result['tp_max_percent']
+                        recover_min_percent = algo_param['recover_min_percent'] if 'recover_min_percent' in algo_param else None
+                        recover_max_pain_percent = algo_param['recover_max_pain_percent'] if 'recover_max_pain_percent' in algo_param else None
+
+                        # tp_min_percent adj: Strategies where target_price not based on tp_max_percent, but variable
+                        if max_pnl_potential_percent and max_pnl_potential_percent<tp_max_percent:
+                            tp_minmax_ratio = tp_min_percent/tp_max_percent
+                            tp_max_percent = max_pnl_potential_percent
+                            tp_min_percent = tp_minmax_ratio * tp_max_percent
+
+                        unrealized_pnl_eval_result = pnl_eval_func(lo_row, lo_row_tm1, running_sl_percent_hard, this_ticker_open_trades, algo_param)
+                        unrealized_pnl_interval = unrealized_pnl_eval_result['unrealized_pnl_interval']
+                        unrealized_pnl_open = unrealized_pnl_eval_result['unrealized_pnl_open']
+                        unrealized_pnl_live_optimistic = unrealized_pnl_eval_result['unrealized_pnl_live_optimistic']
+                        unrealized_pnl_live_pessimistic = unrealized_pnl_eval_result['unrealized_pnl_live_pessimistic']
+                        unrealized_pnl_tp = unrealized_pnl_eval_result['unrealized_pnl_tp']
+                        unrealized_pnl_sl = unrealized_pnl_eval_result['unrealized_pnl_sl']
+                        unrealized_pnl_live = unrealized_pnl_live_pessimistic
+
+                        if unrealized_pnl_live>0 and unrealized_pnl_live_optimistic>max_unrealized_pnl_live:
+                            max_unrealized_pnl_live = unrealized_pnl_live_optimistic
+                            for trade in this_ticker_open_trades:
+                                trade['max_unrealized_pnl_live'] = max_unrealized_pnl_live # Evaluated optimistically!!!
+                        
+                        # Do this before max_pain updated
+                        if unrealized_pnl_live<0 and unrealized_pnl_live_optimistic>max_pain:
+                            recovered_pnl_optimistic = unrealized_pnl_live_optimistic - max_pain
+                            recovered_pnl_pessimistic = unrealized_pnl_live_pessimistic - max_pain
+                            if recovered_pnl_optimistic>max_recovered_pnl:
+                                max_recovered_pnl = recovered_pnl_optimistic
+                                for trade in this_ticker_open_trades:
+                                    trade['max_recovered_pnl'] = max_recovered_pnl
+                                    
+                        if unrealized_pnl_live<0:
+                            if unrealized_pnl_live<max_pain:
+                                max_pain = unrealized_pnl_live
+                                for trade in this_ticker_open_trades:
+                                    trade['max_pain'] = max_pain # unrealized_pnl_live is set to unrealized_pnl_live_pessimistic!
+
+                        if unrealized_pnl_live<0 and max_unrealized_pnl_live>0:
+                            # If out unrealized_pnl_live already fell from positive to negative, reset max_unrealized_pnl_live back to zero
+                            max_unrealized_pnl_live = 0
+                            for trade in this_ticker_open_trades:
+                                trade['max_unrealized_pnl_live'] = max_unrealized_pnl_live
+
+                        unrealized_pnl = unrealized_pnl_live
+                        pnl_percent_notional = unrealized_pnl_open / current_position_usdt * 100 # This is evaluated using open (Don't use close, that's forward bias!)
+                        max_pnl_percent_notional = max_unrealized_pnl_live / current_position_usdt * 100
+                        max_pain_percent_notional = max_pain / current_position_usdt * 100
+                        max_recovered_pnl_percent_notional = max_recovered_pnl / current_position_usdt * 100
+
+                        if (
+                            (pnl_percent_notional>0 and pnl_percent_notional>=tp_min_percent)
+                            or (
+                                recover_max_pain_percent
+                                and pnl_percent_notional<0 
+                                and max_recovered_pnl_percent_notional>=recover_min_percent
+                                and abs(max_pain_percent_notional)>=recover_max_pain_percent
+                            ) # Taking 'abs': Trailing stop can fire if trade moves in either direction - if your trade is losing trade.
+                        ): 
+                            '''
+                            
+                            'effective_tp_trailing_percent' is initialized to float('inf') on entries. Whenever 'pnl_percent_notional' crosses 'tp_min_percent', trailing stop mechanism kicks in.
+
+                            https://norman-lm-fung.medium.com/gradually-tightened-trailing-stops-f7854bf1e02b
+
+                            'effective_tp_trailing_percent' is used to TRIGGER trailing stop.
+                            Please be careful if you're marking closing trade with candle close.
+                            '''
+                            if algo_param['use_gradual_tightened_trailing_stops']:
+                                effective_tp_trailing_percent = calc_eff_trailing_sl(
+                                        tp_min_percent = tp_min_percent,
+                                        tp_max_percent = tp_max_percent,
+                                        sl_percent_trailing = algo_param['sl_percent_trailing'],
+                                        pnl_percent_notional = max_pnl_percent_notional if pnl_percent_notional>0 else max_recovered_pnl_percent_notional,
+                                        default_effective_tp_trailing_percent = float('inf'),
+                                        linear=True if algo_param['trailing_stop_mode']=='linear' else False, # trailing_stop_mode: linear vs parabolic
+                                        pow=5
+                                    )
+                            else:
+                                effective_tp_trailing_percent = algo_param['sl_percent_trailing']
+                        
+                        # 1. SL
+                        if (
+                            unrealized_pnl_live < 0 
+                            or (
+                                (unrealized_pnl_live>0 and unrealized_pnl_live<max_unrealized_pnl_live)
+                                or (
+                                    unrealized_pnl_live<0
+                                    and recovered_pnl_pessimistic<max_recovered_pnl
+                                    and abs(max_recovered_pnl_percent_notional)>=recover_min_percent
+                                    and abs(max_pain_percent_notional)>=recover_max_pain_percent
+                                )
+                            )
+                        ):
+                            # unrealized_pnl_live is set to unrealized_pnl_live_pessimistic!
+                            loss_hard = abs(unrealized_pnl_live)/this_ticker_current_position_usdt * 100 if unrealized_pnl_live<0 else 0
+
+                            if unrealized_pnl_live>0:
+                                loss_trailing = (1 - unrealized_pnl_live/max_unrealized_pnl_live) * 100 if unrealized_pnl_live>0 and unrealized_pnl_live<max_unrealized_pnl_live else 0
+                            elif unrealized_pnl_live<0:
+                                loss_trailing = (1 - recovered_pnl_pessimistic/max_recovered_pnl) * 100 if unrealized_pnl_live<0 and recovered_pnl_pessimistic<max_recovered_pnl else 0
+
+                            if loss_hard>=running_sl_percent_hard:
+                                unrealized_pnl = (running_sl_percent_hard/algo_param['sl_hard_percent']) * unrealized_pnl_sl
+                                reason2 = "sl_hard_percent"
+                            elif (
+                                    loss_trailing>=effective_tp_trailing_percent # loss_trailing is evaluated pessimistically.
+                                    # and pnl_percent_notional>tp_min_percent
+                                    # and unrealized_pnl_live >= sl_trailing_min_threshold_usdt
+                                ):
+                                ''' 
+                                If you're using 'effective_tp_trailing_percent' to approx unrealised pnl, make sure "loss_trailing>=effective_tp_trailing_percent" is the only condition.
+                                Don't AND this with other condition. Otherwise use close price to approx unrealised pnl instead!!!
+                                '''
+                                if unrealized_pnl_live>0:
+                                    unrealized_pnl = min(
+                                        ((100-effective_tp_trailing_percent)/100) * max_unrealized_pnl_live,
+                                        this_ticker_current_position_usdt * algo_param['tp_max_percent']/100
+                                    )
+                                else:
+                                    unrealized_pnl = max_pain + ((100-effective_tp_trailing_percent)/100) * max_recovered_pnl
+                                # unrealized_pnl = unrealized_pnl_interval # less accurate
+                                reason2 = "sl_percent_trailing"
+
+                            if (
+                                (loss_hard>=running_sl_percent_hard) 
+                                or (
+                                    loss_trailing>=effective_tp_trailing_percent 
+                                    # and pnl_percent_notional>tp_min_percent
+                                    # and unrealized_pnl_live >= sl_trailing_min_threshold_usdt
+                                )
+                            ):
+                                block_entry_since_last_sl = True
+                                reason = 'SL' if unrealized_pnl<0 else 'TP'
+                                _close_open_positions(
+                                    key, 
+                                    ticker, 
+                                    this_ticker_current_position_usdt, 
+                                    this_ticker_open_positions_side, 
+                                    current_position_usdt, 
+                                    unrealized_pnl, 
+                                    effective_tp_trailing_percent,
+                                    lo_row, reason, reason2, gloabl_state, all_trades, all_canvas,
+                                    algo_param
+                                )
+                                current_positions_info = _refresh_current_position(lo_timestamp_ms)
+                                avg_entry_price = current_positions_info['avg_entry_price']
+                                pos_side = current_positions_info['side']
+                                current_position_usdt_buy = current_positions_info['current_position_usdt_buy']
+                                current_position_usdt_sell = current_positions_info['current_position_usdt_sell']
+                                current_position_usdt = current_positions_info['current_position_usdt']
+                                this_ticker_open_trades = current_positions_info['this_ticker_open_trades']
+                                this_ticker_current_position_usdt_buy = current_positions_info['this_ticker_current_position_usdt_buy']
+                                this_ticker_current_position_usdt_sell = current_positions_info['this_ticker_current_position_usdt_sell']
+                                this_ticker_open_positions_side = current_positions_info['this_ticker_open_positions_side']
+                                this_ticker_current_position_usdt = current_positions_info['this_ticker_current_position_usdt']
+                                max_sl_trade_age_ms = current_positions_info['max_sl_trade_age_ms']
+
+                                # sl_percent_trailing = algo_param['sl_percent_trailing'] # Reset! Remember!
+                                this_ticker_open_positions_side='flat' # Reset!
+                                reversal_camp_cache[key] = REVERSAL_CAMP_ITEM.copy()
+
+                        # 2. TP: Trigger by unrealized_pnl_live_optimistic, not unrealized_pnl_live (which is unrealized_pnl_live_pessimistic). Pnl estimation from unrealized_pnl_boillenger however!!!
+                        if this_ticker_current_position_usdt>0 and unrealized_pnl_live_optimistic>0:
+                            kwargs = {k: v for k, v in locals().items() if k in tp_eval_func_params}
+                            tp_eval_func_result = tp_eval_func(**kwargs)
+
+                            if tp_eval_func_result:
+                                unrealized_pnl_tp = min(
+                                    unrealized_pnl_tp,
+                                    this_ticker_current_position_usdt * algo_param['tp_max_percent']/100
+                                )
+                                _close_open_positions(
+                                    key, ticker, 
+                                    this_ticker_current_position_usdt, 
+                                    this_ticker_open_positions_side, 
+                                    current_position_usdt, 
+                                    unrealized_pnl_tp, 
+                                    effective_tp_trailing_percent,
+                                    lo_row, 'TP', '', gloabl_state, all_trades, all_canvas,
+                                    algo_param
+                                )
+                                current_position_usdt -= this_ticker_current_position_usdt
+                                this_ticker_current_position_usdt = 0
+                                this_ticker_open_positions_side='flat' # Reset!
+                                reversal_camp_cache[key] = REVERSAL_CAMP_ITEM.copy()
+
+                    def _position_size_and_cash_check(
+                            current_position_usdt : float, # All positions added together (include other tickers)
+                            this_ticker_current_position_usdt : float, # This position only
+                            target_order_notional : float,
+                            total_equity : float,
+                            target_position_size_percent_total_equity : float,
+                            cash : float
+                    ) -> bool:
+                        return (
+                            (current_position_usdt + target_order_notional <= total_equity * (target_position_size_percent_total_equity/100)) 
+                            and (this_ticker_current_position_usdt + target_order_notional <= total_equity * (target_position_size_percent_total_equity/100))
+                            and cash >= target_order_notional
+                        )
+                    
+                    entry_adj_bps = 0 # Essentially disable this for time being.
+
+                    if(
+                        lo_low<=(lo_boillenger_lower*(1-entry_adj_bps/10000))
+                    ):
+                        lo_boillenger_lower_breached_history = lo_boillenger_lower_breached_cache.get(key, [])
+                        lo_boillenger_upper_breached_history = lo_boillenger_upper_breached_cache.get(key, [])
+                        lo_boillenger_lower_breached_cache[key] = lo_boillenger_lower_breached_history
+                        lo_boillenger_upper_breached_cache[key] = lo_boillenger_upper_breached_history
+                        lo_boillenger_upper_breached_history.clear()
+                        lo_boillenger_lower_breached_history.append(lo_datetime)
+                        reversal_camp_cache[key] = REVERSAL_CAMP_ITEM.copy()
+                    elif(
+                        lo_high>=(lo_boillenger_upper*(1+entry_adj_bps/10000))
+                    ):
+                        lo_boillenger_lower_breached_history = lo_boillenger_lower_breached_cache.get(key, [])
+                        lo_boillenger_upper_breached_history = lo_boillenger_upper_breached_cache.get(key, [])
+                        lo_boillenger_lower_breached_cache[key] = lo_boillenger_lower_breached_history
+                        lo_boillenger_upper_breached_cache[key] = lo_boillenger_upper_breached_history
+                        lo_boillenger_lower_breached_history.clear()
+                        lo_boillenger_upper_breached_history.append(lo_datetime)
+                        reversal_camp_cache[key] = REVERSAL_CAMP_ITEM.copy()
+
+                    if algo_param['constant_order_notional']:
+                        target_order_notional = algo_param['target_order_notional']
+                    else:
+                        kwargs = {k: v for k, v in locals().items() if k in order_notional_adj_func_params}
+                        order_notional_adj_func_result = order_notional_adj_func(**kwargs)
+                        target_order_notional = order_notional_adj_func_result['target_order_notional']
+                                    
+                    order_notional_long, order_notional_short = target_order_notional, target_order_notional
+                    if algo_param['clip_order_notional_to_best_volumes']:
+                        order_notional_long = min(lo_volume * lo_low, target_order_notional)
+                        order_notional_short = min(lo_volume * lo_high, target_order_notional)
+                    
+                    kwargs = {k: v for k, v in locals().items() if k in allow_entry_initial_func_params}
+                    allow_entry_initial_func_result = allow_entry_initial_func(**kwargs)
+                    allow_entry_initial_long = allow_entry_initial_func_result['long']
+                    allow_entry_initial_short = allow_entry_initial_func_result['short']
+                    allow_entry_final_long = False
+                    allow_entry_final_short = False
+
+                    if algo_param['enable_sliced_entry']:
+                        kwargs = {k: v for k, v in locals().items() if k in allow_slice_entry_func_params}
+                        allow_slice_entry_func_result = allow_slice_entry_func(**kwargs)
+                        
+                    # 3. Entries
+                    if (
+                            algo_param['strategy_mode'] in [ 'long_only', 'long_short'] 
+                            and order_notional_long>0
+                            and (not algo_param['block_entries_on_impacting_ecoevents'] or num_impacting_economic_calendars==0)
+                            and not block_entry_since_last_sl
+                            and (
+                                (
+                                    this_ticker_open_positions_side=='flat'
+                                    and allow_entry_initial_long
+                                    and _position_size_and_cash_check(current_position_usdt, this_ticker_current_position_usdt, order_notional_long, gloabl_state.total_equity, target_position_size_percent_total_equity, gloabl_state.cash)
+                                ) or (
+                                    this_ticker_open_positions_side=='buy'
+                                    and _position_size_and_cash_check(current_position_usdt, this_ticker_current_position_usdt, order_notional_long, gloabl_state.total_equity, target_position_size_percent_total_equity, gloabl_state.cash)
+                                    and (algo_param['enable_sliced_entry'] and allow_slice_entry_func_result['long'])
+                                )
+                            )
+                    ):
+                        # Long
+                        order_notional = order_notional_long
+
+                        if not reversal_camp_cache[key]['camp1'] and not reversal_camp_cache[key]['camp2'] and not reversal_camp_cache[key]['camp3']:
+                            reversal_camp_cache[key]['camp1'] = True
+                            reversal_camp_cache[key]['camp1_price'] = lo_close
+                        elif reversal_camp_cache[key]['camp1'] and not reversal_camp_cache[key]['camp2'] and not reversal_camp_cache[key]['camp3']:
+                            reversal_camp_cache[key]['camp2'] = True
+                            reversal_camp_cache[key]['camp2_price'] = lo_close
+                        elif reversal_camp_cache[key]['camp1'] and reversal_camp_cache[key]['camp2'] and not reversal_camp_cache[key]['camp3']:
+                            reversal_camp_cache[key]['camp3'] = True
+                            reversal_camp_cache[key]['camp3_price'] = lo_close
+                        reversal_camp_cache[key]['datetime'] = lo_datetime
+
+                        fetch_historical_price_func = fetch_price
+                        kwargs = {k: v for k, v in locals().items() if k in allow_entry_final_func_params}
+                        allow_entry_final_func_result = allow_entry_final_func(**kwargs)
+
+                        allow_entry_final_long = allow_entry_final_func_result['long']
+                        if (allow_entry_final_long):
+                            order_notional_adj_factor = algo_param['dayofweek_adj_map_order_notional'][lo_dayofweek]
+                            if order_notional>0 and order_notional_adj_factor>0:
+                                max_camp = _max_camp(reversal_camp_cache[key]['camp1'], reversal_camp_cache[key]['camp2'], reversal_camp_cache[key]['camp3'])
+                                target_price = allow_entry_final_func_result['target_price_long']
+                                reference_price = allow_entry_final_func_result['reference_price']
+                                sitting_on_boillenger_band = allow_entry_final_func_result['sitting_on_boillenger_band'] if 'sitting_on_boillenger_band' in allow_entry_final_func_result else None
+                                _additional_trade_fields = {k: v for k, v in locals().items() if k in algo_param['additional_trade_fields']}
+                                
+                                commission = order_notional_adj_factor*order_notional * commission_bps / 10000
+                                gloabl_state.total_commission += commission
+
+                                cash_before = gloabl_state.cash
+                                gloabl_state.cash = gloabl_state.cash - order_notional_adj_factor*order_notional - commission
+                                cash_after = gloabl_state.cash
+
+                                running_total_num_positions : int = len([ 1 for x in all_trades if x['reason']=='entry' and not x['closed']])
+
+                                entry_price = allow_entry_final_func_result['entry_price_long']
+                                reversal_camp_cache[key]['price'] = entry_price
+
+                                pnl_potential_bps = (target_price/entry_price - 1) *10000 if target_price else None
+
+                                new_trade_0 = {
+                                                    'trade_datetime' : lo_datetime,
+                                                    'timestamp_ms' : lo_timestamp_ms,
+                                                    'dayofweek' : lo_dayofweek,
+                                                    'exchange' : exchange.name,
+                                                    'symbol' : ticker,
+                                                    'side' : 'buy',
+                                                    'size' : order_notional_adj_factor*order_notional/lo_close, # in base ccy. 
+                                                    'entry_price' : entry_price,
+                                                    'target_price' : target_price,
+                                                    'pnl_potential_bps' : pnl_potential_bps,
+                                                    'ref_ema_close_fast' : ref_ema_close_fast,
+                                                    'running_sl_percent_hard' : algo_param['sl_hard_percent'],
+                                                    'closed' : False,
+                                                    'reason' : 'entry',
+                                                    'reason2' : '',
+                                                    'total_equity' : gloabl_state.total_equity,
+                                                    'this_ticker_current_position_usdt' : this_ticker_current_position_usdt,
+                                                    'current_position_usdt' : current_position_usdt,
+                                                    'running_total_num_positions' : running_total_num_positions,
+                                                    'cash_before' : cash_before,
+                                                    'cash_after' : cash_after,
+                                                    'order_notional' : order_notional,
+                                                    'commission' : commission,
+                                                    'max_pain' : 0,
+                                                    'num_impacting_economic_calendars' : num_impacting_economic_calendars,
+                                                    'max_camp': max_camp,
+                                                    'post_move_price_change_percent' : post_move_price_change_percent
+                                                }
+                                all_trades.append(new_trade_0)
+                                new_trade_0.update(_additional_trade_fields)
+
+                                # Resets!
+                                effective_tp_trailing_percent = float('inf')
+                                lo_boillenger_lower_breached_history = lo_boillenger_lower_breached_cache.get(key, [])
+                                lo_boillenger_lower_breached_history.clear()
+
+                            if plot_timeseries:
+                                '''
+                                https://matplotlib.org/stable/api/_as_gen/matplotlib.axes.Axes.axvline.html
+                                linestyle='-' means solid line. If you don't supply linestyle, the vertical line wont show!!!
+                                '''
+                                all_canvas[f"{key}-param_id{algo_param['param_id']}"]['time_series_canvas'].axvline(x=lo_datetime, color='gray', linewidth=1, linestyle='-')
+                                all_canvas[f"{key}-param_id{algo_param['param_id']}"]['time_series_canvas'].scatter([lo_datetime, lo_datetime], [lo_low, lo_high], color='gray')
+
+                    elif (
+                            algo_param['strategy_mode'] in [ 'short_only', 'long_short'] 
+                            and order_notional_short>0
+                            and (not algo_param['block_entries_on_impacting_ecoevents'] or num_impacting_economic_calendars==0)
+                            and not block_entry_since_last_sl
+                            and (
+                                (
+                                    this_ticker_open_positions_side=='flat'
+                                    and allow_entry_initial_short
+                                    and _position_size_and_cash_check(current_position_usdt, this_ticker_current_position_usdt, order_notional_short, gloabl_state.total_equity, target_position_size_percent_total_equity, gloabl_state.cash)
+                                ) or (
+                                    this_ticker_open_positions_side=='sell'
+                                    and _position_size_and_cash_check(current_position_usdt, this_ticker_current_position_usdt, order_notional_short, gloabl_state.total_equity, target_position_size_percent_total_equity, gloabl_state.cash)
+                                    and (algo_param['enable_sliced_entry'] and allow_slice_entry_func_result['short'])
+                                )
+                            )
+                    ):
+                        # Short
+                        order_notional = order_notional_short
+
+                        if not reversal_camp_cache[key]['camp1'] and not reversal_camp_cache[key]['camp2'] and not reversal_camp_cache[key]['camp3']:
+                            reversal_camp_cache[key]['camp1'] = True
+                            reversal_camp_cache[key]['camp1_price'] = lo_close
+                        elif reversal_camp_cache[key]['camp1'] and not reversal_camp_cache[key]['camp2'] and not reversal_camp_cache[key]['camp3']:
+                            reversal_camp_cache[key]['camp2'] = True
+                            reversal_camp_cache[key]['camp2_price'] = lo_close
+                        elif reversal_camp_cache[key]['camp1'] and reversal_camp_cache[key]['camp2'] and not reversal_camp_cache[key]['camp3']:
+                            reversal_camp_cache[key]['camp3'] = True
+                            reversal_camp_cache[key]['camp3_price'] = lo_close
+                        reversal_camp_cache[key]['datetime'] = lo_datetime
+
+                        fetch_historical_price_func = fetch_price
+                        kwargs = {k: v for k, v in locals().items() if k in allow_entry_final_func_params}
+                        allow_entry_final_func_result = allow_entry_final_func(**kwargs)
+
+                        allow_entry_final_short = allow_entry_final_func_result['short']
+                        if (allow_entry_final_short):
+                            order_notional_adj_factor = algo_param['dayofweek_adj_map_order_notional'][lo_dayofweek]
+                            if order_notional>0 and order_notional_adj_factor>0:
+                                max_camp = _max_camp(reversal_camp_cache[key]['camp1'], reversal_camp_cache[key]['camp2'], reversal_camp_cache[key]['camp3'])
+                                target_price = allow_entry_final_func_result['target_price_short']
+                                reference_price = allow_entry_final_func_result['reference_price']
+                                sitting_on_boillenger_band = allow_entry_final_func_result['sitting_on_boillenger_band'] if 'sitting_on_boillenger_band' in allow_entry_final_func_result else None
+                                _additional_trade_fields = {k: v for k, v in locals().items() if k in algo_param['additional_trade_fields']}
+
+                                commission = order_notional_adj_factor*order_notional * commission_bps / 10000
+                                gloabl_state.total_commission += commission
+
+                                cash_before = gloabl_state.cash
+                                gloabl_state.cash = gloabl_state.cash - order_notional_adj_factor*order_notional - commission
+                                cash_after = gloabl_state.cash
+
+                                running_total_num_positions : int = len([ 1 for x in all_trades if x['reason']=='entry' and not x['closed']])
+
+                                entry_price = allow_entry_final_func_result['entry_price_short']
+                                reversal_camp_cache[key]['price'] = entry_price
+
+                                pnl_potential_bps = (entry_price/target_price - 1) *10000 if target_price else None
+
+                                new_trade_0 = {
+                                                    'trade_datetime' : lo_datetime,
+                                                    'timestamp_ms' : lo_timestamp_ms,
+                                                    'dayofweek' : lo_dayofweek,
+                                                    'exchange' : exchange.name,
+                                                    'symbol' : ticker,
+                                                    'side' : 'sell',
+                                                    'size' : order_notional_adj_factor*order_notional/lo_close, # in base ccy
+                                                    'entry_price' : entry_price,
+                                                    'target_price' : target_price,
+                                                    'pnl_potential_bps' : pnl_potential_bps,
+                                                    'ref_ema_close_fast' : ref_ema_close_fast,
+                                                    'running_sl_percent_hard' : algo_param['sl_hard_percent'],
+                                                    'closed' : False,
+                                                    'reason' : 'entry',
+                                                    'reason2' : '',
+                                                    'total_equity' : gloabl_state.total_equity,
+                                                    'this_ticker_current_position_usdt' : this_ticker_current_position_usdt,
+                                                    'current_position_usdt' : current_position_usdt,
+                                                    'running_total_num_positions' : running_total_num_positions,
+                                                    'cash_before' : cash_before,
+                                                    'cash_after' : cash_after,
+                                                    'order_notional' : order_notional,
+                                                    'commission' : commission,
+                                                    'max_pain' : 0,
+                                                    'num_impacting_economic_calendars' : num_impacting_economic_calendars,
+                                                    'max_camp': max_camp,
+                                                    'post_move_price_change_percent' : post_move_price_change_percent
+                                                }
+                                all_trades.append(new_trade_0)
+                                new_trade_0.update(_additional_trade_fields)
+
+                                # Resets!
+                                effective_tp_trailing_percent = float('inf')
+                                lo_boillenger_upper_breached_history = lo_boillenger_upper_breached_cache.get(key, [])
+                                lo_boillenger_upper_breached_history.clear()
+
+                            if plot_timeseries:
+                                '''
+                                https://matplotlib.org/stable/api/_as_gen/matplotlib.axes.Axes.axvline.html
+                                linestyle='-' means solid line. If you don't supply linestyle, the vertical line wont show!!!
+                                '''
+                                all_canvas[f"{key}-param_id{algo_param['param_id']}"]['time_series_canvas'].axvline(x=lo_datetime, color='gray', linewidth=1, linestyle='-')
+                                all_canvas[f"{key}-param_id{algo_param['param_id']}"]['time_series_canvas'].scatter([lo_datetime, lo_datetime], [lo_low, lo_high], color='gray')
+                    
+                    iter_info = f"param_id: {algo_param['param_id']}, {key} i: {i} {lo_datetime}, # trades: {len(all_trades)}, equity: {round(gloabl_state.total_equity,2)}"
+                    if i%100==0 and i%1000!=0:
+                        print(iter_info)
+                    elif i%1000==0:
+                        logger.info(iter_info)
+                    
+                if i==pd_lo_candles.shape[0]-1:
+                    # HC
+                    if this_ticker_current_position_usdt>0:
+                        _close_open_positions(key, ticker, this_ticker_current_position_usdt, this_ticker_open_positions_side, current_position_usdt, unrealized_pnl, None, lo_row, 'HC', '', gloabl_state, all_trades, all_canvas, algo_param)
+                    
+            sorted_filtered_tickers.clear()
+            sorted_filtered_tickers = None
+
+        if gloabl_state.total_equity<target_order_notional:
+            logger.warning(f"total_equity {gloabl_state.total_equity} < target_order_notional {target_order_notional} exiting prematurely on {lo_datetime}!!!")
+            break
+
+    if plot_timeseries:
+        for exchange in exchanges:
+            for ticker in tickers:
+                key = f"{exchange.name}-{ticker}-param_id{algo_param['param_id']}"
+                canvas = all_canvas[key]
+                canvas['plt'].savefig(f"ts_{key.replace('/','').replace(':','')}.jpg", format='jpg', dpi=300)
+
+    for reference_price_cache_file in reference_price_cache:
+        reference_price_cache[reference_price_cache_file].sort_values("timestamp_ms", inplace=True)
+        reference_price_cache[reference_price_cache_file].to_csv(reference_price_cache_file)
+
+    num_tp = len([ x for x in all_trades if x['reason']=='TP'])
+    num_sl = len([ x for x in all_trades if x['reason']=='SL'])
+    num_hc_tp = len([ x for x in all_trades if x['reason']=='HC' and x['trade_pnl']>0 ] )
+    num_hc_sl = len([ x for x in all_trades if x['reason']=='HC' and x['trade_pnl']<=0 ] )
+    num_hc = num_hc_tp + num_hc_sl
+                 
+    return {
+        'realized_pnl' : sum([x['trade_pnl'] for x in all_trades if 'trade_pnl' in x]) - gloabl_state.total_commission,
+        'total_commission' : gloabl_state.total_commission,
+        'hit_ratio' : (num_tp + num_hc_tp) / (num_tp + num_sl + num_hc),
+        'num_tp' : num_tp,
+        'num_sl' : num_sl,
+        'num_hc' : num_hc,
+        'num_entry' : num_tp + num_sl + num_hc,
+        'trades' : all_trades,
+        'exceptions' : exceptions
+    }
+
+def run_all_scenario(
+    algo_params : List[Dict[str, Any]],
+    exchanges : List[Exchange],
+
+    order_notional_adj_func : Callable[..., float],
+    allow_entry_initial_func : Callable[..., bool],
+    allow_entry_final_func : Callable[..., bool],
+    allow_slice_entry_func : Callable[..., bool],
+    sl_adj_func : Callable[..., Dict[str, float]],
+    trailing_stop_threshold_eval_func : Callable[..., Dict[str, float]],
+    pnl_eval_func : Callable[..., Dict[str, float]],
+    tp_eval_func : Callable[..., bool],
+    sort_filter_universe_func : Callable[..., List[str]],
+
+    logger,
+
+    reference_start_dt : datetime = datetime(2021,1,1, tzinfo=timezone.utc),
+) -> List[Dict]:
+    all_exceptions = []
+
+    start = datetime.now()
+    max_test_end_date = start
+
+    economic_calendars_file = algo_params[0]['economic_calendars_file']
+    ecoevents_mapped_regions = algo_params[0]['ecoevents_mapped_regions']
+    pd_economic_calendars = None
+    economic_calendars_loaded : bool = False
+    if os.path.isfile(economic_calendars_file):
+        pd_economic_calendars  = pd.read_csv(economic_calendars_file)
+        pd_economic_calendars = pd_economic_calendars[pd_economic_calendars.region.isin(ecoevents_mapped_regions)]
+        economic_calendars_loaded = True if pd_economic_calendars.shape[0]>0 else False
+
+    i : int = 1
+    algo_results : List[Dict] = []
+    best_realized_pnl, best_algo_result = 0, None
+    for algo_param in algo_params:
+
+        algo_result : Dict = {
+            'param' : algo_param
+        }
+        algo_results.append(algo_result)
+
+        # We calc test_end_date with 'lo' (not with 'hi'), we assume it'd be the same.
+        test_start_date = algo_param['start_date']
+        test_fetch_start_date = test_start_date
+        lo_candle_size = algo_param['lo_candle_size']
+        lo_num_intervals = int(lo_candle_size[0])
+        lo_interval = lo_candle_size[-1]
+        lo_how_many_candles = algo_param['lo_how_many_candles']
+        if lo_interval=="m":
+            test_end_date = test_start_date + timedelta(minutes=lo_num_intervals*lo_how_many_candles) 
+            test_fetch_start_date = test_fetch_start_date - timedelta(minutes=algo_param['lo_stats_computed_over_how_many_candles']*2) 
+            test_end_date_ref = test_end_date + timedelta(minutes=algo_param['lo_stats_computed_over_how_many_candles']*4) 
+        elif lo_interval=="h":
+            test_end_date = test_start_date + timedelta(hours=lo_num_intervals*lo_how_many_candles) 
+            test_fetch_start_date = test_fetch_start_date - timedelta(hours=algo_param['lo_stats_computed_over_how_many_candles']*2) 
+            test_end_date_ref = test_end_date + timedelta(hours=algo_param['lo_stats_computed_over_how_many_candles']*4) 
+        elif lo_interval=="d":
+            test_end_date = test_start_date + timedelta(days=lo_num_intervals*lo_how_many_candles) 
+            test_fetch_start_date = test_fetch_start_date - timedelta(days=algo_param['lo_stats_computed_over_how_many_candles']*2) 
+            test_end_date_ref = test_end_date + timedelta(days=algo_param['lo_stats_computed_over_how_many_candles']*4) 
+        test_end_date = test_end_date if test_end_date < max_test_end_date else max_test_end_date
+        test_end_date_ref = test_end_date_ref if test_end_date_ref < max_test_end_date else max_test_end_date
+        cutoff_ts = int(test_fetch_start_date.timestamp()) # in seconds
+        
+
+        ####################################### STEP 1. Fetch candles (Because each test may have diff test_end_date, you need re-fetch candles for each algo_param) #######################################
+        '''
+        cutoff_ts in seconds, example '1668135382'
+
+        exchanges[0].fetch_ohlcv('ETHUSDT', "1m", cutoff_ts)
+
+        Candles format, first field is timestamp in ms:
+            [
+                [1502942400000, 301.13, 301.13, 301.13, 301.13, 0.42643],
+                [1502942460000, 301.13, 301.13, 301.13, 301.13, 2.75787],
+                [1502942520000, 300.0, 300.0, 300.0, 300.0, 0.0993],
+                [1502942580000, 300.0, 300.0, 300.0, 300.0, 0.31389],
+                ...
+            ]
+        '''
+        delisted : List[str] = []
+
+        data_fetch_start : float = time.time()
+        
+        # Fetch BTC
+        reference_ticker : str = algo_param['reference_ticker']
+        target_candle_file_name_fast : str = f'{reference_ticker.replace("^","").replace("/","").replace(":","")}_fast_candles_{datetime(2021,1,1, tzinfo=timezone.utc).strftime("%Y-%m-%d-%H-%M-%S")}_{test_end_date_ref.strftime("%Y-%m-%d-%H-%M-%S")}_1d.csv'
+        target_candle_file_name_slow : str = f'{reference_ticker.replace("^","").replace("/","").replace(":","")}_slow_candles_{datetime(2021,1,1, tzinfo=timezone.utc).strftime("%Y-%m-%d-%H-%M-%S")}_{test_end_date_ref.strftime("%Y-%m-%d-%H-%M-%S")}_1d.csv'
+        logger.info(f"reference_ticker: {reference_ticker}, target_candle_file_name_fast: {target_candle_file_name_fast}, target_candle_file_name_slow: {target_candle_file_name_slow}, reference_candles_file: {algo_param['reference_candles_file'] if 'reference_candles_file' in algo_param else '---'}")
+        if algo_param['force_reload'] or not os.path.isfile(target_candle_file_name_fast):
+            if algo_param['force_reload'] and 'reference_candles_file' in algo_param and algo_param['reference_candles_file'] and os.path.isfile(algo_param['reference_candles_file']):
+                pd_ref_candles_fast = pd.read_csv(algo_param['reference_candles_file'])
+                pd_ref_candles_slow : pd.DataFrame = pd_ref_candles_fast.copy(deep=True)
+                logger.info(f"reference candles loaded from {algo_param['reference_candles_file']}")
+
+            else:
+                ref_candles : Dict[str, pd.DataFrame] = fetch_candles(
+                                                                                        start_ts=int(reference_start_dt.timestamp()), 
+                                                                                        end_ts=int(test_end_date_ref.timestamp()), 
+                                                                                        exchange=exchanges[0], 
+                                                                                        normalized_symbols=[reference_ticker], 
+                                                                                        candle_size = '1d', 
+                                                                                        num_candles_limit=algo_param['num_candles_limit'],
+                                                                                        logger=logger,
+                                                                                        cache_dir=algo_param['cache_candles'],
+                                                                                        list_ts_field=exchanges[0].options['list_ts_field'] if 'list_ts_field' in exchanges[0].options else None
+                                                                                        )
+                logger.info(f"Reference candles fetched: {reference_ticker}, start: {reference_start_dt}, end: {test_end_date_ref}")
+                pd_ref_candles_fast : pd.DataFrame = ref_candles[reference_ticker]
+                pd_ref_candles_slow : pd.DataFrame = pd_ref_candles_fast.copy(deep=True)
+
+            compute_candles_stats(pd_candles=pd_ref_candles_fast, boillenger_std_multiples=2, sliding_window_how_many_candles=algo_param['ref_ema_num_days_fast'], slow_fast_interval_ratio=int(algo_param['ref_ema_num_days_fast']/2), rsi_sliding_window_how_many_candles=algo_param['rsi_sliding_window_how_many_candles'], rsi_trend_sliding_window_how_many_candles=algo_param['rsi_trend_sliding_window_how_many_candles'], hurst_exp_window_how_many_candles=algo_param['hurst_exp_window_how_many_candles'], target_fib_level=algo_param['target_fib_level'], pypy_compat=algo_param['pypy_compat'])
+            compute_candles_stats(pd_candles=pd_ref_candles_slow, boillenger_std_multiples=2, sliding_window_how_many_candles=algo_param['ref_ema_num_days_slow'], slow_fast_interval_ratio=int(algo_param['ref_ema_num_days_slow']/2), rsi_sliding_window_how_many_candles=algo_param['rsi_sliding_window_how_many_candles'], rsi_trend_sliding_window_how_many_candles=algo_param['rsi_trend_sliding_window_how_many_candles'], hurst_exp_window_how_many_candles=algo_param['hurst_exp_window_how_many_candles'], target_fib_level=algo_param['target_fib_level'], pypy_compat=algo_param['pypy_compat'])
+            logger.info(f"Reference candles {reference_ticker} compute_candles_stats done.")
+            
+            pd_ref_candles_fast.to_csv(target_candle_file_name_fast)
+            pd_ref_candles_slow.to_csv(target_candle_file_name_slow)
+
+        else:
+            pd_ref_candles_fast : pd.DataFrame = pd.read_csv(target_candle_file_name_fast)
+            pd_ref_candles_slow : pd.DataFrame = pd.read_csv(target_candle_file_name_slow)
+            fix_column_types(pd_ref_candles_fast)
+            fix_column_types(pd_ref_candles_slow)
+            logger.info(f"Reference candles {reference_ticker} loaded from target_candle_file_name_fast: {target_candle_file_name_fast}, target_candle_file_name_slow: {target_candle_file_name_slow}")
+
+        total_seconds = (test_end_date_ref - test_start_date).total_seconds()
+        total_hours = total_seconds / 3600
+        total_days = total_hours / 24
+        sliding_window_how_many_candles : int = 0
+        sliding_window_how_many_candles = int(total_days / algo_param['sliding_window_ratio'])
+
+        ref_candles_partitions, pd_hi_candles_partitions, pd_lo_candles_partitions = None, None, None
+        if not algo_param['pypy_compat']:
+            ref_candles_partitions = partition_sliding_window(
+                    pd_candles = pd_ref_candles_fast, 
+                    sliding_window_how_many_candles = sliding_window_how_many_candles, 
+                    smoothing_window_size_ratio = algo_param['smoothing_window_size_ratio'], 
+                    linregress_stderr_threshold = algo_param['linregress_stderr_threshold'],
+                    max_recur_depth = algo_param['max_recur_depth'], 
+                    min_segment_size_how_many_candles = algo_param['min_segment_size_how_many_candles'], 
+                    segment_consolidate_slope_ratio_threshold = algo_param['segment_consolidate_slope_ratio_threshold'],
+                    sideway_price_condition_threshold = algo_param['sideway_price_condition_threshold']
+                )
+            candle_segments_jpg_file_name : str = f'{reference_ticker.replace("^","").replace("/","").replace(":","")}_refcandles_w_segments_{datetime(2021,1,1, tzinfo=timezone.utc).strftime("%Y-%m-%d-%H-%M-%S")}_{test_end_date_ref.strftime("%Y-%m-%d-%H-%M-%S")}_1d.jpg'
+            plot_segments(pd_ref_candles_fast, ref_candles_partitions, candle_segments_jpg_file_name)
+
+            candle_segments_file_name : str = f'{reference_ticker.replace("^","").replace("/","").replace(":","")}_refcandles_w_segments_{datetime(2021,1,1, tzinfo=timezone.utc).strftime("%Y-%m-%d-%H-%M-%S")}_{test_end_date_ref.strftime("%Y-%m-%d-%H-%M-%S")}_1d.csv'
+            pd_ref_candles_segments = segments_to_df(ref_candles_partitions['segments'])
+            pd_ref_candles_segments.to_csv(candle_segments_file_name)
+        
+        all_exchange_candles : Dict[str, Dict[str, Dict[str, pd.DataFrame]]] = {}
+        for exchange in exchanges:
+            markets = exchange.load_markets()
+            if exchange.name not in all_exchange_candles:
+                all_exchange_candles[exchange.name]  = {} 
+
+                if algo_param['white_list_tickers']:
+                    tickers = algo_param['white_list_tickers']
+                else:
+                    tickers = list(markets.keys())
+
+                for ticker in tickers:
+                    if ticker not in markets:
+                        err_msg = f"{ticker}: {'no longer in markets'}"
+                        logger.error(err_msg)
+                        delisted.append(ticker)
+                    else:    
+                        all_exchange_candles[exchange.name][ticker] = {}
+
+                        _ticker = ticker.split(":")[0].replace("/","")
+                        total_seconds = (test_end_date - test_fetch_start_date).total_seconds()
+                        total_hours = total_seconds / 3600
+                        total_days = total_hours / 24
+                        sliding_window_how_many_candles : int = 0
+                        sliding_window_how_many_candles = int(total_days / algo_param['sliding_window_ratio'])
+                    
+                        pd_hi_candles = None
+                        target_candle_file_name : str = f'{_ticker}_candles_{test_fetch_start_date.strftime("%Y-%m-%d-%H-%M-%S")}_{test_end_date.strftime("%Y-%m-%d-%H-%M-%S")}_{algo_param["hi_candle_size"]}.csv'
+                        if algo_param['force_reload'] or not os.path.isfile(target_candle_file_name):
+                            if algo_param['force_reload'] and 'hi_candles_file' in algo_param and algo_param['hi_candles_file'] and os.path.isfile(algo_param['hi_candles_file']):
+                                pd_hi_candles : pd.DataFrame = pd.read_csv(algo_param['hi_candles_file'])
+
+                            else:
+                                hi_candles : Dict[str, pd.DataFrame] = fetch_candles(
+                                                                                                start_ts=cutoff_ts, 
+                                                                                                end_ts=int(test_end_date.timestamp()), 
+                                                                                                exchange=exchange, normalized_symbols=[ ticker ], 
+                                                                                                candle_size = algo_param['hi_candle_size'], 
+                                                                                                num_candles_limit=algo_param['num_candles_limit'],
+                                                                                                logger=logger,
+                                                                                                cache_dir=algo_param['cache_candles'],
+                                                                                                list_ts_field=exchange.options['list_ts_field']
+                                                                                                )
+                                pd_hi_candles : pd.DataFrame = hi_candles[ticker]
+                                logger.info(f"pd_hi_candles fetched: {ticker} {pd_hi_candles.shape}, start: {cutoff_ts}, end: {int(test_end_date.timestamp())}")
+                            compute_candles_stats(pd_candles=pd_hi_candles, boillenger_std_multiples=algo_param['boillenger_std_multiples'], sliding_window_how_many_candles=algo_param['hi_stats_computed_over_how_many_candles'], slow_fast_interval_ratio=(algo_param['hi_stats_computed_over_how_many_candles']/algo_param['hi_ma_short_interval']), rsi_sliding_window_how_many_candles=algo_param['rsi_sliding_window_how_many_candles'], rsi_trend_sliding_window_how_many_candles=algo_param['rsi_trend_sliding_window_how_many_candles'], hurst_exp_window_how_many_candles=algo_param['hurst_exp_window_how_many_candles'], target_fib_level=algo_param['target_fib_level'], pypy_compat=algo_param['pypy_compat'])
+                            logger.info(f"pd_hi_candles {ticker} compute_candles_stats done: {target_candle_file_name}")
+                            pd_hi_candles.to_csv(target_candle_file_name)
+
+                            if pd_hi_candles is not None and pd_hi_candles.shape[0]>0:
+                                first_candle_datetime = datetime.fromtimestamp(pd_hi_candles.iloc[0]['timestamp_ms']/1000)
+                                last_candle_datetime = datetime.fromtimestamp(pd_hi_candles.iloc[-1]['timestamp_ms']/1000)
+
+                                assert(last_candle_datetime>first_candle_datetime)
+                            else:
+                                err_msg = f"{ticker} no hi candles?"
+                                logger.error(err_msg)
+                        else:
+                            pd_hi_candles : pd.DataFrame = pd.read_csv(target_candle_file_name)
+                            fix_column_types(pd_hi_candles)
+                            logger.info(f"pd_hi_candles {ticker} {pd_hi_candles.shape} loaded from {target_candle_file_name}")
+
+                        if not algo_param['pypy_compat']:
+                            pd_hi_candles_partitions = partition_sliding_window(
+                                        pd_candles = pd_hi_candles, 
+                                        sliding_window_how_many_candles = sliding_window_how_many_candles, 
+                                        smoothing_window_size_ratio = algo_param['smoothing_window_size_ratio'], 
+                                        linregress_stderr_threshold = algo_param['linregress_stderr_threshold'],
+                                        max_recur_depth = algo_param['max_recur_depth'], 
+                                        min_segment_size_how_many_candles = algo_param['min_segment_size_how_many_candles'], 
+                                        segment_consolidate_slope_ratio_threshold = algo_param['segment_consolidate_slope_ratio_threshold'],
+                                        sideway_price_condition_threshold = algo_param['sideway_price_condition_threshold']
+                                    )
+                            candle_segments_jpg_file_name : str = f'{_ticker}_hicandles_w_segments_{test_fetch_start_date.strftime("%Y-%m-%d-%H-%M-%S")}_{test_end_date.strftime("%Y-%m-%d-%H-%M-%S")}_{algo_param["hi_candle_size"]}.jpg'
+                            plot_segments(pd_hi_candles, pd_hi_candles_partitions, candle_segments_jpg_file_name)
+
+                            candle_segments_file_name : str = f'{_ticker}_hicandles_w_segments_{test_fetch_start_date.strftime("%Y-%m-%d-%H-%M-%S")}_{test_end_date.strftime("%Y-%m-%d-%H-%M-%S")}_{algo_param["hi_candle_size"]}.csv'
+                            pd_hi_candles_segments = segments_to_df(pd_hi_candles_partitions['segments'])
+                            pd_hi_candles_segments.to_csv(candle_segments_file_name)
+
+                        pd_lo_candles = None
+                        _ticker = ticker.split(":")[0].replace("/","")
+                        target_candle_file_name : str = f'{_ticker}_candles_{test_fetch_start_date.strftime("%Y-%m-%d-%H-%M-%S")}_{test_end_date.strftime("%Y-%m-%d-%H-%M-%S")}_{algo_param["lo_candle_size"]}.csv'
+                        if algo_param['force_reload'] or not os.path.isfile(target_candle_file_name):
+                            if algo_param['force_reload'] and 'lo_candles_file' in algo_param and algo_param['lo_candles_file'] and os.path.isfile(algo_param['lo_candles_file']):
+                                pd_lo_candles : pd.DataFrame = pd.read_csv(algo_param['lo_candles_file'])
+
+                            else:
+                                lo_candles : Dict[str, pd.DataFrame] = fetch_candles(
+                                                                                                start_ts=cutoff_ts, 
+                                                                                                end_ts=int(test_end_date.timestamp()), 
+                                                                                                exchange=exchange, normalized_symbols=[ ticker ], 
+                                                                                                candle_size = algo_param['lo_candle_size'], 
+                                                                                                num_candles_limit=algo_param['num_candles_limit'],
+                                                                                                logger=logger,
+                                                                                                cache_dir=algo_param['cache_candles'],
+                                                                                                list_ts_field=exchange.options['list_ts_field']
+                                                                                                )
+                                pd_lo_candles : pd.DataFrame = lo_candles[ticker]
+                                logger.info(f"pd_lo_candles fetched: {ticker} {pd_lo_candles.shape}, start: {cutoff_ts}, end: {int(test_end_date.timestamp())}")
+                            compute_candles_stats(pd_candles=pd_lo_candles, boillenger_std_multiples=algo_param['boillenger_std_multiples'], sliding_window_how_many_candles=algo_param['lo_stats_computed_over_how_many_candles'], slow_fast_interval_ratio=(algo_param['lo_stats_computed_over_how_many_candles']/algo_param['lo_ma_short_interval']), rsi_sliding_window_how_many_candles=algo_param['rsi_sliding_window_how_many_candles'], rsi_trend_sliding_window_how_many_candles=algo_param['rsi_trend_sliding_window_how_many_candles'], hurst_exp_window_how_many_candles=algo_param['hurst_exp_window_how_many_candles'], target_fib_level=algo_param['target_fib_level'], pypy_compat=algo_param['pypy_compat'])
+                            logger.info(f"pd_lo_candles {ticker} compute_candles_stats done. {target_candle_file_name}")
+                            pd_lo_candles.to_csv(target_candle_file_name)
+                            
+                            if pd_lo_candles is not None and pd_lo_candles.shape[0]>0:
+                                first_candle_datetime = datetime.fromtimestamp(pd_lo_candles.iloc[0]['timestamp_ms']/1000)
+                                last_candle_datetime = datetime.fromtimestamp(pd_lo_candles.iloc[-1]['timestamp_ms']/1000)
+                                
+                                assert(last_candle_datetime>first_candle_datetime)
+                            else:
+                                err_msg = f"{ticker} no lo candles?"
+                                logger.error(err_msg)
+                        else:
+                            pd_lo_candles : pd.DataFrame = pd.read_csv(target_candle_file_name)
+                            fix_column_types(pd_lo_candles)
+                            logger.info(f"pd_lo_candles {ticker} {pd_lo_candles.shape} loaded from {target_candle_file_name}")
+
+                        if not algo_param['pypy_compat']:
+                            pd_lo_candles_partitions = partition_sliding_window(
+                                        pd_candles = pd_lo_candles, 
+                                        sliding_window_how_many_candles = sliding_window_how_many_candles, 
+                                        smoothing_window_size_ratio = algo_param['smoothing_window_size_ratio'], 
+                                        linregress_stderr_threshold = algo_param['linregress_stderr_threshold'],
+                                        max_recur_depth = algo_param['max_recur_depth'], 
+                                        min_segment_size_how_many_candles = algo_param['min_segment_size_how_many_candles'], 
+                                        segment_consolidate_slope_ratio_threshold = algo_param['segment_consolidate_slope_ratio_threshold'],
+                                        sideway_price_condition_threshold = algo_param['sideway_price_condition_threshold']
+                                    )
+                            candle_segments_jpg_file_name : str = f'{_ticker}_locandles_w_segments_{test_fetch_start_date.strftime("%Y-%m-%d-%H-%M-%S")}_{test_end_date.strftime("%Y-%m-%d-%H-%M-%S")}_{algo_param["lo_candle_size"]}.jpg'
+                            plot_segments(pd_lo_candles, pd_lo_candles_partitions, candle_segments_jpg_file_name)
+
+                            candle_segments_file_name : str = f'{_ticker}_locandles_w_segments_{test_fetch_start_date.strftime("%Y-%m-%d-%H-%M-%S")}_{test_end_date.strftime("%Y-%m-%d-%H-%M-%S")}_{algo_param["hi_candle_size"]}.csv'
+                            pd_lo_candles_segments = segments_to_df(pd_lo_candles_partitions['segments'])
+                            pd_lo_candles_segments.to_csv(candle_segments_file_name)
+                    
+                        all_exchange_candles[exchange.name][ticker]['hi_candles'] = pd_hi_candles
+                        all_exchange_candles[exchange.name][ticker]['lo_candles'] = pd_lo_candles
+
+        data_fetch_finish : float = time.time()
+
+        ####################################### STEP 2. Trade simulation #######################################
+        logger.info(f"Start run_scenario")
+        scenario_start : float = time.time()
+        result = run_scenario(
+                algo_param=algo_param, 
+                exchanges=exchanges, 
+                all_exchange_candles=all_exchange_candles, 
+                pd_ref_candles_fast=pd_ref_candles_fast, 
+                pd_ref_candles_slow=pd_ref_candles_slow, 
+                ref_candles_partitions=ref_candles_partitions,
+                pd_hi_candles_partitions=pd_hi_candles_partitions,
+                pd_lo_candles_partitions=pd_lo_candles_partitions,
+                economic_calendars_loaded=economic_calendars_loaded,
+                pd_economic_calendars=pd_economic_calendars,
+                tickers=tickers, 
+
+                order_notional_adj_func=order_notional_adj_func,
+                allow_entry_initial_func=allow_entry_initial_func,
+                allow_entry_final_func=allow_entry_final_func,
+                allow_slice_entry_func=allow_slice_entry_func,
+                sl_adj_func=sl_adj_func,
+                trailing_stop_threshold_eval_func=trailing_stop_threshold_eval_func,
+                pnl_eval_func=pnl_eval_func,
+                tp_eval_func=tp_eval_func,
+                sort_filter_universe_func=sort_filter_universe_func,
+
+                logger=logger,
+
+                pypy_compat=algo_param['pypy_compat'],
+                plot_timeseries=True
+            )
+        scenario_finish = time.time()
+        
+        data_fetch_elapsed_ms = (data_fetch_finish - data_fetch_start) * 1000
+        scenario_elapsed_ms = (scenario_finish - scenario_start) * 1000
+
+        logger.info(f"Done run_scenario. data_fetch_elapsed_ms: {data_fetch_elapsed_ms} ms, scenario_elapsed_ms: {scenario_elapsed_ms}")
+        
+        algo_result['orders'] = result['trades']
+        result.pop('trades')
+        algo_result['summary'] = {
+            # Key parameters
+            'initial_cash' : algo_param['initial_cash'],
+            'entry_percent_initial_cash' : algo_param['entry_percent_initial_cash'],
+            'strategy_mode' : algo_param['strategy_mode'],
+            'ref_ema_num_days_fast' : algo_param['ref_ema_num_days_fast'],
+            'ref_ema_num_days_slow' : algo_param['ref_ema_num_days_slow'],
+            'long_above_ref_ema_short_below' : algo_param['long_above_ref_ema_short_below'],
+            'ref_price_vs_ema_percent_threshold' : algo_param['ref_price_vs_ema_percent_threshold'] if 'ref_price_vs_ema_percent_threshold' in algo_param else None,
+            'rsi_upper_threshold' : algo_param['rsi_upper_threshold'],
+            'rsi_lower_threshold' : algo_param['rsi_lower_threshold'],
+            'boillenger_std_multiples' : algo_param['boillenger_std_multiples'],
+            'ema_short_slope_threshold' : algo_param['ema_short_slope_threshold'] if 'ema_short_slope_threshold' in algo_param else None,
+            'num_intervals_block_pending_ecoevents' : algo_param['num_intervals_block_pending_ecoevents'],
+            'num_intervals_current_ecoevents' : algo_param['num_intervals_current_ecoevents'],
+            'sl_hard_percent' : algo_param['sl_hard_percent'],
+            'sl_percent_trailing' : algo_param['sl_percent_trailing'],
+            'use_gradual_tightened_trailing_stops' : algo_param['use_gradual_tightened_trailing_stops'],
+            'sl_num_intervals_delay' : algo_param['sl_num_intervals_delay'],
+            'tp_min_percent' : algo_param['tp_min_percent'],
+            'tp_max_percent' : algo_param['tp_max_percent'],
+            'asymmetric_tp_bps' : algo_param['asymmetric_tp_bps'],
+
+            # Key output
+            'realized_pnl' : result['realized_pnl'], # Commission already taken out
+            'total_commission' : result['total_commission'],
+            'hit_ratio' : result['hit_ratio'],
+            'num_tp' : result['num_tp'],
+            'num_sl' : result['num_sl'],
+            'num_hc' : result['num_hc'],
+            'num_entry' : result['num_entry'],
+            'data_fetch_elapsed_ms' : data_fetch_elapsed_ms,
+            'scenario_elapsed_ms' : scenario_elapsed_ms,
+            'num_exceptions' : len(result['exceptions'])
+        }
+
+        all_exceptions = all_exceptions + list(result['exceptions'].items())
+        logger.error(list(result['exceptions'].items()))
+
+        logger.info(f"Done ({i}/{len(algo_params)}) {algo_param['name_exclude_start_date']}")
+        logger.info(json.dumps(algo_result['summary'], indent=4))
+
+        if result['realized_pnl']>best_realized_pnl or not best_algo_result:
+            best_algo_result = algo_result['summary']
+        
+        i = i + 1
+
+    finish = datetime.now()
+    elapsed = (finish-start).seconds
+
+
+    logger.info(f"Backtest done in {elapsed}sec over {len(algo_params)} scenario's with start_date {test_start_date} over {len(exchanges)} exchange(s) and {len(tickers)} tickers.")
+
+    logger.info(f"*** Best result realized_pnl: {best_algo_result['realized_pnl']}")
+    logger.info(json.dumps(best_algo_result, indent=4))
+
+    pd_results = pd.DataFrame([ x['summary'] for x in algo_results])
+    pd_results.loc['avg', 'realized_pnl'] = pd_results['realized_pnl'].mean(numeric_only=True, axis=0)
+    pd_results.loc['avg', 'total_commission'] = pd_results['total_commission'].mean(numeric_only=True, axis=0)
+    pd_results.loc['avg', 'hit_ratio'] = pd_results['hit_ratio'].mean(numeric_only=True, axis=0)
+
+    return algo_results
+
+def parseargs():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--force_reload", help="Reload candles? Both candles and TA previously computed will be loaded from disk. Y or N (default)", default=False)
+    parser.add_argument("--white_list_tickers", help="Comma seperated list, example: BTC/USDT:USDT,ETH/USDT:USDT,XRP/USDT:USDT ", default="BTC/USDT:USDT")
+    parser.add_argument("--reference_ticker", help="This is ticker for bull / bear determination. The Northstar.", default="BTC/USDT:USDT")
+    parser.add_argument("--block_entries_on_impacting_ecoevents", help="Block entries on economic event? Y (default) or N", default=True)
+    parser.add_argument("--enable_sliced_entry", help="Block entries on economic event? Y or N (default)", default=False)
+    parser.add_argument("--asymmetric_tp_bps", help="A positive asymmetric_tp_bps means you are taking deeper TPs. A negative asymmetric_tp_bps means shallower", default=0)
+    args = parser.parse_args()
+    
+    if args.force_reload:
+        if args.force_reload=='Y':
+            force_reload = True
+        else:
+            force_reload = False
+    else:
+        force_reload = False
+
+    if args.white_list_tickers:
+        white_list_tickers = args.white_list_tickers.split(',')
+        
+    reference_ticker = args.reference_ticker if args.reference_ticker else white_list_tickers[0]
+
+    if args.block_entries_on_impacting_ecoevents:
+        if args.block_entries_on_impacting_ecoevents=='Y':
+            block_entries_on_impacting_ecoevents = True
+        else:
+            block_entries_on_impacting_ecoevents = False
+    else:
+        block_entries_on_impacting_ecoevents = True
+
+    if args.enable_sliced_entry:
+        if args.enable_sliced_entry=='Y':
+            enable_sliced_entry = True
+        else:
+            enable_sliced_entry = False
+    else:
+        enable_sliced_entry = False
+
+    asymmetric_tp_bps = int(args.asymmetric_tp_bps)
+
+    return {
+        'force_reload': force_reload,
+        'white_list_tickers' : white_list_tickers,
+        'reference_ticker' : reference_ticker,
+        'block_entries_on_impacting_ecoevents' : block_entries_on_impacting_ecoevents,
+        'enable_sliced_entry'  : enable_sliced_entry,
+        'asymmetric_tp_bps' : asymmetric_tp_bps
+    }
+
+def dump_trades_to_disk(
+    algo_results,
+    filename,
+    logger
+):
+    flattenned_trades : List[Dict[str, Any]]= []
+    for algo_result in algo_results:
+        for order in algo_result['orders']:
+            try:
+                order['name'] = algo_result['param']['name']
+                order['name_exclude_start_date'] = algo_result['param']['name_exclude_start_date']
+                
+                order['initial_cash'] = algo_result['param']['initial_cash']
+                order['entry_percent_initial_cash'] = algo_result['param']['entry_percent_initial_cash']
+                order['clip_order_notional_to_best_volumes'] = algo_result['param']['clip_order_notional_to_best_volumes']
+                order['target_position_size_percent_total_equity'] = algo_result['param']['target_position_size_percent_total_equity']
+
+                order['reference_ticker'] = algo_result['param']['reference_ticker']
+                order['strategy_mode'] = algo_result['param']['strategy_mode']
+                order['boillenger_std_multiples'] = algo_result['param']['boillenger_std_multiples']
+                order['ema_short_slope_threshold'] = algo_result['param']['ema_short_slope_threshold'] if 'ema_short_slope_threshold' in algo_result['param'] else None
+                order['how_many_last_candles'] = algo_result['param']['how_many_last_candles']
+                order['last_candles_timeframe'] = algo_result['param']['last_candles_timeframe']
+                order['enable_wait_entry'] = algo_result['param']['enable_wait_entry'] if 'enable_wait_entry' in algo_result['param'] else None
+                order['allow_entry_sit_bb'] = algo_result['param']['allow_entry_sit_bb']  if 'allow_entry_sit_bb' in algo_result['param'] else None
+                order['enable_sliced_entry'] = algo_result['param']['enable_sliced_entry']
+                order['adj_sl_on_ecoevents'] = algo_result['param']['adj_sl_on_ecoevents']
+                order['block_entries_on_impacting_ecoevents'] = algo_result['param']['block_entries_on_impacting_ecoevents']
+                order['num_intervals_block_pending_ecoevents'] = algo_result['param']['num_intervals_block_pending_ecoevents']
+                order['num_intervals_current_ecoevents'] = algo_result['param']['num_intervals_current_ecoevents']
+                order['enable_hi_timeframe_confirm'] = algo_result['param']['enable_hi_timeframe_confirm'] if 'enable_hi_timeframe_confirm' in algo_result['param'] else None
+                order['sl_num_intervals_delay'] = algo_result['param']['sl_num_intervals_delay']
+                order['sl_hard_percent'] = algo_result['param']['sl_hard_percent']
+                order['sl_percent_trailing'] = algo_result['param']['sl_percent_trailing']
+                order['use_gradual_tightened_trailing_stops'] = algo_result['param']['use_gradual_tightened_trailing_stops']
+                order['tp_min_percent'] = algo_result['param']['tp_min_percent']
+                order['tp_max_percent'] = algo_result['param']['tp_max_percent']
+                order['asymmetric_tp_bps'] = algo_result['param']['asymmetric_tp_bps']
+            
+                order['hi_candle_size'] = algo_result['param']['hi_candle_size']
+                order['hi_stats_computed_over_how_many_candles'] = algo_result['param']['hi_stats_computed_over_how_many_candles']
+                order['hi_how_many_candles'] = algo_result['param']['hi_how_many_candles']
+                order['hi_ma_short_interval'] = algo_result['param']['hi_ma_short_interval']
+                order['hi_ma_long_interval'] = algo_result['param']['hi_ma_long_interval']
+
+                order['lo_candle_size'] = algo_result['param']['lo_candle_size']
+                order['lo_stats_computed_over_how_many_candles'] = algo_result['param']['lo_stats_computed_over_how_many_candles']
+                order['lo_how_many_candles'] = algo_result['param']['lo_how_many_candles']
+                order['lo_ma_short_interval'] = algo_result['param']['lo_ma_short_interval']
+                order['lo_ma_long_interval'] = algo_result['param']['lo_ma_long_interval']
+
+                order['target_fib_level'] = algo_result['param']['target_fib_level']
+                order['rsi_sliding_window_how_many_candles'] = algo_result['param']['rsi_sliding_window_how_many_candles']
+                order['rsi_trend_sliding_window_how_many_candles'] = algo_result['param']['rsi_trend_sliding_window_how_many_candles']
+                order['hurst_exp_window_how_many_candles'] = algo_result['param']['hurst_exp_window_how_many_candles']
+
+                order['ref_ema_num_days_fast'] = algo_result['param']['ref_ema_num_days_fast']
+                order['re_ema_num_days_slow'] = algo_result['param']['ref_ema_num_days_slow']
+                order['long_above_ref_ema_short_below'] = algo_result['param']['long_above_ref_ema_short_below']
+                order['ref_price_vs_ema_percent_threshold'] = algo_result['param']['ref_price_vs_ema_percent_threshold'] if 'ref_price_vs_ema_percent_threshold' in algo_result['param'] else None
+                order['rsi_upper_threshold'] = algo_result['param']['rsi_upper_threshold']
+                order['rsi_lower_threshold'] = algo_result['param']['rsi_lower_threshold']
+
+                order['id'] = str(uuid.uuid4())
+                order['trade_year'] = order['trade_datetime'].year
+                order['trade_month'] = order['trade_datetime'].month
+                order['trade_day'] = order['trade_datetime'].day
+                order['trade_dayofweek'] = order['dayofweek']
+                order['trade_week_of_month'] = timestamp_to_week_of_month(
+                    int(order['trade_datetime'].timestamp() * 1000)
+                )
+                
+                flattenned_trades.append(order)
+
+            except Exception as error:
+                logger.error(f"Error while processing flattenned trades! {error}")
+
+    if len(flattenned_trades)>0:
+        pd_flattenned_trades = pd.DataFrame(flattenned_trades)
+        pd_flattenned_trades.to_csv(filename)
+
+        logger.info(f"Trade extract: {filename}")