aria-code 0.167.0__py3-none-win_amd64.whl

aria_code/domain/backtest_report.py

@@ -0,0 +1,488 @@
+"""Local real-data strategy backtest and self-contained HTML rendering."""
+
+from __future__ import annotations
+
+import html
+import math
+import re
+from dataclasses import dataclass
+from datetime import date, datetime, timedelta
+from pathlib import Path
+from typing import Any, Dict, Iterable, List, Optional, Sequence, Tuple
+
+from aria_code.artifacts import create_user_artifact, write_artifact_metadata, write_artifact_raw_data
+from aria_code.data_service import DataService
+from aria_code.domain import performance_stats as stats
+
+
+@dataclass(frozen=True)
+class BacktestConfig:
+    symbol: str
+    strategy: str = "momentum"
+    start_date: Optional[str] = None
+    end_date: Optional[str] = None
+    initial_capital: float = 100000.0
+    fast_period: int = 20
+    slow_period: int = 60
+    momentum_period: int = 20
+
+
+def _as_float(value: Any) -> Optional[float]:
+    try:
+        if value is None:
+            return None
+        out = float(value)
+        return out if math.isfinite(out) else None
+    except Exception:
+        return None
+
+
+def _parse_date(value: Any) -> Optional[date]:
+    if not value:
+        return None
+    try:
+        return datetime.fromisoformat(str(value)[:10]).date()
+    except Exception:
+        return None
+
+
+def _history_days(start_date: Optional[str], end_date: Optional[str]) -> int:
+    end = _parse_date(end_date) or date.today()
+    start = _parse_date(start_date)
+    if not start:
+        return 365
+    return max((end - start).days + 10, 90)
+
+
+def _clean_history_rows(rows: Iterable[Dict[str, Any]], start_date: Optional[str], end_date: Optional[str]) -> List[Dict[str, Any]]:
+    start = _parse_date(start_date)
+    end = _parse_date(end_date)
+    cleaned: List[Dict[str, Any]] = []
+    for row in rows or []:
+        row_date = _parse_date(row.get("date") or row.get("Date"))
+        close = _as_float(row.get("close", row.get("Close")))
+        if not row_date or close is None or close <= 0:
+            continue
+        if start and row_date < start:
+            continue
+        if end and row_date > end:
+            continue
+        cleaned.append(
+            {
+                "date": row_date.isoformat(),
+                "close": close,
+                "open": _as_float(row.get("open", row.get("Open"))),
+                "high": _as_float(row.get("high", row.get("High"))),
+                "low": _as_float(row.get("low", row.get("Low"))),
+                "volume": _as_float(row.get("volume", row.get("Volume"))),
+            }
+        )
+    cleaned.sort(key=lambda x: x["date"])
+    return cleaned
+
+
+def _sma(values: Sequence[float], window: int) -> List[Optional[float]]:
+    if window <= 1:
+        return [float(v) for v in values]
+    out: List[Optional[float]] = []
+    rolling = 0.0
+    for i, value in enumerate(values):
+        rolling += value
+        if i >= window:
+            rolling -= values[i - window]
+        out.append(rolling / window if i + 1 >= window else None)
+    return out
+
+
+# Every name accepted, mapped to the strategy it runs.
+STRATEGY_ALIASES = {
+    "buy_hold": "buy_hold", "buyhold": "buy_hold", "hold": "buy_hold", "buy_and_hold": "buy_hold",
+    "sma_cross": "sma_cross", "ma_cross": "sma_cross", "moving_average": "sma_cross", "sma": "sma_cross",
+    "sma_crossover": "sma_cross", "ma_crossover": "sma_cross",
+    "momentum": "momentum", "mom": "momentum",
+}
+
+
+def normalize_strategy(name: str) -> Optional[str]:
+    """The strategy a name runs, or None if there is no such strategy."""
+    return STRATEGY_ALIASES.get((name or "momentum").lower().replace("-", "_").replace(" ", "_"))
+
+
+def _signals(strategy: str, closes: Sequence[float], fast: int, slow: int, momentum_period: int) -> List[int]:
+    known = normalize_strategy(strategy)
+    if known is None:
+        # Any other name used to fall through to buy-and-hold and be reported
+        # under the name asked for: "/backtest sma AAPL" was buy-and-hold.
+        raise ValueError(f"Unknown strategy '{strategy}'. Available: momentum, sma_cross (sma), buy_hold")
+    strategy = known
+    n = len(closes)
+    if strategy == "buy_hold":
+        return [1] * n
+    if strategy == "sma_cross":
+        fast_ma = _sma(closes, max(2, fast))
+        slow_ma = _sma(closes, max(max(3, slow), fast + 1))
+        return [1 if f is not None and s is not None and f > s else 0 for f, s in zip(fast_ma, slow_ma)]
+    period = max(2, momentum_period)
+    return [1 if i >= period and closes[i] > closes[i - period] else 0 for i in range(n)]
+
+
+def run_backtest_from_history(history: Sequence[Dict[str, Any]], config: BacktestConfig) -> Dict[str, Any]:
+    rows = _clean_history_rows(history, config.start_date, config.end_date)
+    min_bars = max(30, min(max(config.slow_period, config.momentum_period) + 2, 80))
+    if len(rows) < min_bars:
+        return {
+            "success": False,
+            "symbol": config.symbol,
+            "error": f"历史行情不足：需要至少 {min_bars} 根K线，当前 {len(rows)} 根",
+        }
+
+    dates = [str(r["date"]) for r in rows]
+    closes = [float(r["close"]) for r in rows]
+    volumes = [_as_float(r.get("volume")) for r in rows]
+    valid_volumes = [v for v in volumes if v is not None and v >= 0]
+    try:
+        signals = _signals(config.strategy, closes, config.fast_period, config.slow_period, config.momentum_period)
+    except ValueError as exc:
+        return {"success": False, "symbol": config.symbol, "strategy": config.strategy, "error": str(exc)}
+
+    initial = float(config.initial_capital or 100000.0)
+    equity = [initial]
+    benchmark = [initial]
+    daily_strategy_returns = [0.0]
+    daily_benchmark_returns = [0.0]
+    trades = 0
+    previous_position = 0
+    trade_results: List[float] = []   # return of each round trip
+    entry_equity = None
+
+    for i in range(1, len(closes)):
+        day_return = closes[i] / closes[i - 1] - 1.0
+        position = signals[i - 1]  # shift one day to avoid look-ahead bias
+        if position == 1 and previous_position == 0:
+            trades += 1
+            entry_equity = equity[-1]
+        elif position == 0 and previous_position == 1 and entry_equity:
+            trade_results.append(equity[-1] / entry_equity - 1.0)
+            entry_equity = None
+        previous_position = position
+        strategy_return = day_return * position
+        daily_strategy_returns.append(strategy_return)
+        daily_benchmark_returns.append(day_return)
+        equity.append(equity[-1] * (1.0 + strategy_return))
+        benchmark.append(initial * closes[i] / closes[0])
+
+    total_return = equity[-1] / initial - 1.0
+    benchmark_return = benchmark[-1] / initial - 1.0
+    rets = daily_strategy_returns[1:]
+    years = stats.span_years(dates[0], dates[-1])
+    if years is None:
+        # Undatable rows: count bars as days, the long-standing fallback.
+        years = max(len(rows) / stats.DAYS_PER_YEAR, stats.MIN_YEARS)
+        periods_per_year = float(stats.FALLBACK_PERIODS_PER_YEAR)
+    else:
+        periods_per_year = len(rets) / years
+    annualized_return = stats.annualized_return(total_return, years)
+    volatility = stats.stddev(rets) * math.sqrt(periods_per_year)
+    sharpe = stats.sharpe(rets, periods_per_year)
+    if entry_equity:
+        trade_results.append(equity[-1] / entry_equity - 1.0)   # still open: marked at the last close
+    # Winning trades over trades. This was the share of up days while holding,
+    # so one buy-and-hold "trade" showed a 52.5% win rate.
+    win_rate = sum(1 for r in trade_results if r > 0) / len(trade_results) if trade_results else 0.0
+
+    curve = [
+        {
+            "date": d,
+            "strategy": round(e, 4),
+            "benchmark": round(b, 4),
+            "close": round(c, 4),
+            "position": int(sig),
+        }
+        for d, e, b, c, sig in zip(dates, equity, benchmark, closes, signals)
+    ]
+
+    return {
+        "success": True,
+        "symbol": config.symbol.upper(),
+        "strategy": config.strategy,
+        "start": dates[0],
+        "end": dates[-1],
+        "bars": len(rows),
+        "initial_capital": initial,
+        "total_return": round(total_return, 6),
+        "annualized_return": round(annualized_return, 6),
+        "annual_return": round(annualized_return, 6),
+        "benchmark_return": round(benchmark_return, 6),
+        "buy_hold_return": round(benchmark_return, 6),
+        "alpha": round(total_return - benchmark_return, 6),
+        "max_drawdown": round(stats.max_drawdown(equity), 6),
+        "benchmark_max_drawdown": round(stats.max_drawdown(benchmark), 6),
+        "annual_volatility": round(volatility, 6),
+        "sharpe_ratio": round(sharpe, 4),
+        "win_rate": round(win_rate, 4),
+        "total_trades": trades,
+        "volume_summary": {
+            "last": round(valid_volumes[-1], 2) if valid_volumes else None,
+            "average": round(sum(valid_volumes) / len(valid_volumes), 2) if valid_volumes else None,
+            "min": round(min(valid_volumes), 2) if valid_volumes else None,
+            "max": round(max(valid_volumes), 2) if valid_volumes else None,
+            "coverage": round(len(valid_volumes) / len(rows), 4) if rows else 0.0,
+        },
+        "equity_curve": curve,
+    }
+
+
+def _fmt_pct(value: Any) -> str:
+    number = _as_float(value)
+    if number is None:
+        return "-"
+    return f"{number * 100:+.2f}%"
+
+
+def _fmt_num(value: Any, digits: int = 2) -> str:
+    number = _as_float(value)
+    if number is None:
+        return "-"
+    return f"{number:,.{digits}f}"
+
+
+def _points(values: Sequence[float], width: int, height: int, pad: int) -> str:
+    if not values:
+        return ""
+    lo = min(values)
+    hi = max(values)
+    if math.isclose(lo, hi):
+        lo *= 0.99
+        hi *= 1.01
+    usable_w = max(width - pad * 2, 1)
+    usable_h = max(height - pad * 2, 1)
+    pts = []
+    for i, value in enumerate(values):
+        x = pad + usable_w * (i / max(len(values) - 1, 1))
+        y = pad + usable_h * (1 - (value - lo) / (hi - lo))
+        pts.append(f"{x:.1f},{y:.1f}")
+    return " ".join(pts)
+
+
+def render_backtest_html(result: Dict[str, Any], output_path: Path) -> Path:
+    output_path.parent.mkdir(parents=True, exist_ok=True)
+    curve = result.get("equity_curve") or []
+    strategy_values = [_as_float(p.get("strategy")) or 0.0 for p in curve if isinstance(p, dict)]
+    benchmark_values = [_as_float(p.get("benchmark")) or 0.0 for p in curve if isinstance(p, dict)]
+    width, height, pad = 920, 360, 34
+    strategy_points = _points(strategy_values, width, height, pad)
+    benchmark_points = _points(benchmark_values, width, height, pad)
+    start = html.escape(str(result.get("start", "")))
+    end = html.escape(str(result.get("end", "")))
+    symbol = html.escape(str(result.get("symbol", "")))
+    strategy = html.escape(str(result.get("strategy", "")))
+    created_at = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+    bars = int(result.get("bars") or len(curve))
+    latest_strategy = strategy_values[-1] if strategy_values else 0.0
+    latest_benchmark = benchmark_values[-1] if benchmark_values else 0.0
+    provider_chain = [str(p) for p in (result.get("provider_chain") or []) if p]
+    missing_fields = [str(p) for p in (result.get("missing_fields") or []) if p]
+    data_status = html.escape(str(result.get("data_status") or "unknown"))
+    data_provider = html.escape(str(result.get("data_provider") or "history"))
+    data_updated_at = html.escape(str(result.get("data_updated_at") or ""))
+    source_text = html.escape(" → ".join(provider_chain) if provider_chain else data_provider)
+    missing_text = html.escape(", ".join(missing_fields) if missing_fields else "none")
+
+    metrics = [
+        ("策略收益", _fmt_pct(result.get("total_return"))),
+        ("买入持有", _fmt_pct(result.get("benchmark_return"))),
+        ("超额收益", _fmt_pct(result.get("alpha"))),
+        ("年化收益", _fmt_pct(result.get("annualized_return"))),
+        ("最大回撤", _fmt_pct(result.get("max_drawdown"))),
+        ("夏普比率", _fmt_num(result.get("sharpe_ratio"), 2)),
+        ("胜率", _fmt_pct(result.get("win_rate"))),
+        ("交易次数", str(result.get("total_trades", 0))),
+    ]
+    metric_html = "\n".join(
+        f"<div class=\"metric\"><span>{html.escape(k)}</span><strong>{html.escape(v)}</strong></div>"
+        for k, v in metrics
+    )
+
+    html_doc = f"""<!doctype html>
+<html lang="zh-CN">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1">
+<title>{symbol} {strategy} Backtest</title>
+<style>
+:root {{
+  color-scheme: light dark;
+  --bg: #f7f7f4;
+  --panel: #ffffff;
+  --text: #1f1f1f;
+  --muted: #787878;
+  --line: #d8d4ce;
+  --accent: #b8794b;
+  --bench: #6d6d6d;
+}}
+@media (prefers-color-scheme: dark) {{
+  :root {{
+    --bg: #181818;
+    --panel: #222222;
+    --text: #eeeeee;
+    --muted: #9a9a9a;
+    --line: #3a3a3a;
+    --accent: #d08a52;
+    --bench: #aaaaaa;
+  }}
+}}
+* {{ box-sizing: border-box; }}
+body {{
+  margin: 0;
+  background: var(--bg);
+  color: var(--text);
+  font-family: ui-monospace, SFMono-Regular, Menlo, Consolas, monospace;
+  line-height: 1.45;
+}}
+main {{ max-width: 1080px; margin: 0 auto; padding: 28px 20px 40px; }}
+.top {{ display: flex; justify-content: space-between; gap: 16px; align-items: baseline; margin-bottom: 18px; }}
+h1 {{ font-size: 24px; margin: 0; letter-spacing: 0; }}
+.sub {{ color: var(--muted); font-size: 14px; }}
+.panel {{ background: var(--panel); border: 1px solid var(--line); border-radius: 8px; padding: 18px; margin-top: 14px; }}
+.metrics {{ display: grid; grid-template-columns: repeat(auto-fit, minmax(150px, 1fr)); gap: 10px; }}
+.metric {{ border-top: 1px solid var(--line); padding-top: 10px; min-height: 58px; }}
+.metric span {{ display: block; color: var(--muted); font-size: 12px; margin-bottom: 4px; }}
+.metric strong {{ font-size: 18px; }}
+.legend {{ display: flex; gap: 18px; color: var(--muted); font-size: 13px; margin-top: 12px; }}
+.legend i {{ display: inline-block; width: 18px; height: 3px; vertical-align: middle; margin-right: 6px; }}
+.provenance {{ display: grid; grid-template-columns: repeat(auto-fit, minmax(190px, 1fr)); gap: 10px; font-size: 13px; }}
+.provenance div {{ border-top: 1px solid var(--line); padding-top: 8px; }}
+.provenance span {{ display: block; color: var(--muted); font-size: 12px; margin-bottom: 3px; }}
+svg {{ width: 100%; height: auto; display: block; }}
+.foot {{ color: var(--muted); font-size: 12px; margin-top: 14px; }}
+</style>
+</head>
+<body>
+<main>
+  <div class="top">
+    <div>
+      <h1>{symbol} · {strategy} 策略回测</h1>
+      <div class="sub">{start} → {end} · {bars} bars · real historical data</div>
+    </div>
+    <div class="sub">Aria Code · {html.escape(created_at)}</div>
+  </div>
+  <section class="panel metrics">
+    {metric_html}
+  </section>
+  <section class="panel">
+    <svg viewBox="0 0 {width} {height}" role="img" aria-label="equity curve">
+      <rect x="0" y="0" width="{width}" height="{height}" fill="transparent"/>
+      <line x1="{pad}" y1="{height - pad}" x2="{width - pad}" y2="{height - pad}" stroke="var(--line)"/>
+      <line x1="{pad}" y1="{pad}" x2="{pad}" y2="{height - pad}" stroke="var(--line)"/>
+      <polyline points="{benchmark_points}" fill="none" stroke="var(--bench)" stroke-width="2" stroke-dasharray="6 6"/>
+      <polyline points="{strategy_points}" fill="none" stroke="var(--accent)" stroke-width="3"/>
+      <text x="{pad}" y="{pad - 10}" fill="var(--muted)" font-size="12">策略权益 {html.escape(_fmt_num(latest_strategy, 0))}</text>
+      <text x="{width - pad - 210}" y="{pad - 10}" fill="var(--muted)" font-size="12">基准权益 {html.escape(_fmt_num(latest_benchmark, 0))}</text>
+    </svg>
+    <div class="legend">
+      <span><i style="background:var(--accent)"></i>Strategy equity</span>
+      <span><i style="background:var(--bench)"></i>Buy & hold</span>
+    </div>
+  </section>
+  <section class="panel provenance">
+    <div><span>Data status</span>{data_status}</div>
+    <div><span>Provider chain</span>{source_text}</div>
+    <div><span>Missing fields</span>{missing_text}</div>
+    <div><span>Updated at</span>{data_updated_at or html.escape(created_at)}</div>
+  </section>
+  <div class="foot">本报告为本地生成的历史模拟，不构成投资建议。信号按前一交易日收盘后生效，避免未来函数。</div>
+</main>
+</body>
+</html>
+"""
+    output_path.write_text(html_doc, encoding="utf-8")
+    return output_path
+
+
+def generate_backtest_report(
+    config: BacktestConfig,
+    output_dir: Optional[Path] = None,
+    market_client: Optional[Any] = None,
+) -> Dict[str, Any]:
+    data_service = DataService(market_client=market_client) if market_client is not None else DataService()
+    days = _history_days(config.start_date, config.end_date)
+    hist_result = data_service.history(config.symbol, days=days, interval="1d")
+    hist = hist_result.data
+    if not hist_result.success:
+        return {
+            "success": False,
+            "symbol": config.symbol,
+            "strategy": config.strategy,
+            "error": hist.get("friendly_error") or hist.get("error") or "历史行情获取失败",
+            "provider_chain": hist_result.provider_chain,
+            "missing_fields": hist_result.missing_fields,
+            "data_status": "data_unavailable",
+            "data_warnings": hist_result.warnings,
+        }
+
+    result = run_backtest_from_history(hist.get("data") or [], config)
+    result["data_provider"] = hist_result.source or hist.get("provider")
+    result["provider_chain"] = hist_result.provider_chain or hist.get("provider_chain") or [hist.get("provider", "history")]
+    result["missing_fields"] = hist_result.missing_fields
+    result["data_warnings"] = hist_result.warnings
+    result["data_status"] = "complete" if hist_result.success and not hist_result.missing_fields else "partial"
+    result["data_updated_at"] = hist_result.timestamp
+    if not result.get("success"):
+        return result
+
+    safe_symbol = re.sub(r"[^A-Za-z0-9_.-]+", "_", config.symbol.upper()).strip("_") or "SYMBOL"
+    safe_strategy = re.sub(r"[^A-Za-z0-9_.-]+", "_", config.strategy.lower()).strip("_") or "strategy"
+    ts_dt = datetime.now()
+    if output_dir:
+        root = Path(output_dir)
+        root.mkdir(parents=True, exist_ok=True)
+        output_path = root / f"{safe_symbol}_{safe_strategy}_{ts_dt.strftime('%Y%m%d_%H%M%S')}.html"
+        artifact = None
+    else:
+        artifact = create_user_artifact(
+            "strategies/backtests",
+            config.symbol,
+            f"{safe_symbol}_{safe_strategy}_backtest",
+            ".html",
+            timestamp=ts_dt,
+        )
+        output_path = artifact.path
+    render_backtest_html(result, output_path)
+    if artifact:
+        write_artifact_metadata(artifact, {
+            "kind": "strategy_backtest",
+            "status": "complete",
+            "symbol": config.symbol,
+            "strategy": config.strategy,
+            "created_at": ts_dt.isoformat(timespec="seconds"),
+            "data": {
+                "provider": result.get("data_provider"),
+                "provider_chain": result.get("provider_chain"),
+                "missing_fields": result.get("missing_fields"),
+                "status": result.get("data_status"),
+                "updated_at": result.get("data_updated_at"),
+                "rows": len(hist.get("data") or []),
+            },
+            "config": config.__dict__,
+            "metrics": {
+                key: result.get(key)
+                for key in ("total_return", "annual_return", "sharpe", "max_drawdown", "win_rate", "trades")
+            },
+        })
+        write_artifact_raw_data(artifact, {
+            "symbol": config.symbol,
+            "strategy": config.strategy,
+            "history": hist.get("data") or [],
+            "data": {
+                "provider": result.get("data_provider"),
+                "provider_chain": result.get("provider_chain"),
+                "missing_fields": result.get("missing_fields"),
+                "status": result.get("data_status"),
+                "warnings": result.get("data_warnings"),
+                "updated_at": result.get("data_updated_at"),
+            },
+            "result": result,
+        })
+    result["report_path"] = str(output_path)
+    result["provider"] = "local_backtest"
+    return result