hockey-blast-common-lib 0.1.63__py3-none-any.whl → 0.1.64__py3-none-any.whl

hockey_blast_common_lib/skills_propagation.py

@@ -1,19 +1,23 @@
-import sys
 import os
+import sys
 from collections import defaultdict
+
 import numpy as np
 
 # Add the project root directory to the Python path
 sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
 
-from hockey_blast_common_lib.models import Level, Division
-from hockey_blast_common_lib.stats_models import LevelsGraphEdge, LevelStatsSkater, SkillValuePPGRatio
+from sqlalchemy.exc import IntegrityError
+
 from hockey_blast_common_lib.db_connection import create_session
+from hockey_blast_common_lib.models import Division, Level
 from hockey_blast_common_lib.progress_utils import create_progress_tracker
-from sqlalchemy import func
-from sqlalchemy.exc import IntegrityError
+from hockey_blast_common_lib.stats_models import (
+    LevelsGraphEdge,
+    LevelStatsSkater,
+    SkillValuePPGRatio,
+)
 
-import numpy as np
 
 class Config:
     MIN_GAMES_PLAYED_FOR_EDGE = 10
@@ -42,6 +46,7 @@ class Config:
         modified_z_score = 0.6745 * diff / med_abs_deviation
         return data[modified_z_score < m]
 
+
 def reset_skill_values_in_divisions():
     session = create_session("boss")
 
@@ -52,14 +57,18 @@ def reset_skill_values_in_divisions():
         # Look up the Skill table using the level from Division
         div_level = division.level
         # Query to find the matching Skill
-        level = session.query(Level).filter(Level.org_id == division.org_id, Level.level_name == div_level).one_or_none()
+        level = (
+            session.query(Level)
+            .filter(Level.org_id == division.org_id, Level.level_name == div_level)
+            .one_or_none()
+        )
 
         if not level:
             # If no match found, check each alternative name individually
             skills = session.query(Level).filter(Level.org_id == division.org_id).all()
             for s in skills:
                 if s.level_alternative_name:  # Check if not None
-                    alternative_names = s.level_alternative_name.split(',')
+                    alternative_names = s.level_alternative_name.split(",")
                     if div_level in alternative_names:
                         level = s
                         break
@@ -74,11 +83,14 @@ def reset_skill_values_in_divisions():
                 level.skill_value = -1
         else:
             # Check if level already exists with this org_id/level_name combination
-            existing_level = session.query(Level).filter(
-                Level.org_id == division.org_id,
-                Level.level_name == division.level
-            ).first()
-            
+            existing_level = (
+                session.query(Level)
+                .filter(
+                    Level.org_id == division.org_id, Level.level_name == division.level
+                )
+                .first()
+            )
+
             if existing_level:
                 # Use existing level
                 division.level_id = existing_level.id
@@ -89,9 +101,9 @@ def reset_skill_values_in_divisions():
                     org_id=division.org_id,
                     skill_value=-1,
                     level_name=division.level,
-                    level_alternative_name='',
+                    level_alternative_name="",
                     is_seed=False,
-                    skill_propagation_sequence=-1
+                    skill_propagation_sequence=-1,
                 )
                 session.add(new_level)
                 try:
@@ -101,20 +113,31 @@ def reset_skill_values_in_divisions():
                 except IntegrityError:
                     session.rollback()
                     # Another process created this level, query for it
-                    existing_level = session.query(Level).filter(
-                        Level.org_id == division.org_id,
-                        Level.level_name == division.level
-                    ).first()
+                    existing_level = (
+                        session.query(Level)
+                        .filter(
+                            Level.org_id == division.org_id,
+                            Level.level_name == division.level,
+                        )
+                        .first()
+                    )
                     if existing_level:
                         division.level_id = existing_level.id
-                        print(f"Race condition resolved - using existing Level for Division {division.level}")
+                        print(
+                            f"Race condition resolved - using existing Level for Division {division.level}"
+                        )
                     else:
-                        raise RuntimeError(f"Unable to create or find level: {division.level} for org_id: {division.org_id}")
+                        raise RuntimeError(
+                            f"Unable to create or find level: {division.level} for org_id: {division.org_id}"
+                        )
 
         # Commit the changes to the Division
         session.commit()
 
-    print("Level values and propagation sequences have been populated into the Division table.")
+    print(
+        "Level values and propagation sequences have been populated into the Division table."
+    )
+
 
 def build_levels_graph_edges():
     # Creates unique edges from levelA to levelB (there is no reverse edge levelB to levelA)
@@ -131,10 +154,13 @@ def build_levels_graph_edges():
     level_human_stats = defaultdict(lambda: defaultdict(dict))
 
     for stat in level_stats:
-        if stat.games_played >= Config.MIN_GAMES_PLAYED_FOR_EDGE and stat.points_per_game >= Config.MIN_PPG_FOR_EDGE:
+        if (
+            stat.games_played >= Config.MIN_GAMES_PLAYED_FOR_EDGE
+            and stat.points_per_game >= Config.MIN_PPG_FOR_EDGE
+        ):
             level_human_stats[stat.level_id][stat.human_id] = {
-                'games_played': stat.games_played,
-                'points_per_game': stat.points_per_game
+                "games_played": stat.games_played,
+                "points_per_game": stat.points_per_game,
             }
 
     # Dictionary to store edges
@@ -142,8 +168,11 @@ def build_levels_graph_edges():
 
     # Build edges - batch load all levels first for performance
     all_level_ids = list(level_human_stats.keys())
-    levels_dict = {level.id: level for level in session.query(Level).filter(Level.id.in_(all_level_ids)).all()}
-    
+    levels_dict = {
+        level.id: level
+        for level in session.query(Level).filter(Level.id.in_(all_level_ids)).all()
+    }
+
     total_levels = len(level_human_stats)
     progress = create_progress_tracker(total_levels, "Building level graph edges")
     processed_levels = 0
@@ -167,10 +196,10 @@ def build_levels_graph_edges():
             # if from_level.id == 223 and to_level.id == 219: #216
             #     print(f"Debug: From Level ID: {from_level.id}, To Level ID: {to_level.id}")
             for human_id in common_humans:
-                from_ppg = from_humans[human_id]['points_per_game']
-                to_ppg = to_humans[human_id]['points_per_game']
-                from_games = from_humans[human_id]['games_played']
-                to_games = to_humans[human_id]['games_played']
+                from_ppg = from_humans[human_id]["points_per_game"]
+                to_ppg = to_humans[human_id]["points_per_game"]
+                from_games = from_humans[human_id]["games_played"]
+                to_games = to_humans[human_id]["games_played"]
                 min_games = min(from_games, to_games)
                 n_games += min_games
 
@@ -199,7 +228,7 @@ def build_levels_graph_edges():
                 to_level_id=to_level_id,
                 n_connections=n_connections,
                 ppg_ratio=avg_ppg_ratio,
-                n_games=n_games  # Store the number of games
+                n_games=n_games,  # Store the number of games
             )
             edges[(from_level_id, to_level_id)] = edge
 
@@ -213,9 +242,10 @@ def build_levels_graph_edges():
 
     print("\nLevels graph edges have been populated into the database.")
 
+
 def propagate_skill_levels(propagation_sequence):
-    min_skill_value = float('inf')
-    max_skill_value = float('-inf')
+    min_skill_value = float("inf")
+    max_skill_value = float("-inf")
 
     session = create_session("boss")
 
@@ -225,7 +255,9 @@ def propagate_skill_levels(propagation_sequence):
         session.commit()
 
         # Build and save the correlation data
-        levels = session.query(Level).filter(Level.skill_propagation_sequence == 0).all()
+        levels = (
+            session.query(Level).filter(Level.skill_propagation_sequence == 0).all()
+        )
         level_ids = {level.id for level in levels}
         correlation_data = defaultdict(list)
 
@@ -233,10 +265,14 @@ def propagate_skill_levels(propagation_sequence):
             if level.skill_value == -1:
                 continue
 
-            edges = session.query(LevelsGraphEdge).filter(
-                (LevelsGraphEdge.from_level_id == level.id) |
-                (LevelsGraphEdge.to_level_id == level.id)
-            ).all()
+            edges = (
+                session.query(LevelsGraphEdge)
+                .filter(
+                    (LevelsGraphEdge.from_level_id == level.id)
+                    | (LevelsGraphEdge.to_level_id == level.id)
+                )
+                .all()
+            )
 
             for edge in edges:
                 if edge.n_connections < Config.MIN_CONNECTIONS_FOR_CORRELATION:
@@ -252,7 +288,9 @@ def propagate_skill_levels(propagation_sequence):
                 if target_level_id not in level_ids:
                     continue
 
-                target_level = session.query(Level).filter_by(id=target_level_id).first()
+                target_level = (
+                    session.query(Level).filter_by(id=target_level_id).first()
+                )
                 if target_level:
                     skill_value_from = level.skill_value
                     skill_value_to = target_level.skill_value
@@ -261,10 +299,12 @@ def propagate_skill_levels(propagation_sequence):
                     if skill_value_from == skill_value_to:
                         continue
 
-
                     # Since we go over all levels in the sequence 0, we will see each edge twice
                     # This condition eliminates duplicates
-                    if abs(skill_value_from - skill_value_to) > Config.MAX_SKILL_DIFF_IN_EDGE:
+                    if (
+                        abs(skill_value_from - skill_value_to)
+                        > Config.MAX_SKILL_DIFF_IN_EDGE
+                    ):
                         continue
 
                     # Debug prints
@@ -272,7 +312,10 @@ def propagate_skill_levels(propagation_sequence):
 
                     # Ensure INCREASING SKILL VALUES for the correlation data!
                     if skill_value_from > skill_value_to:
-                        skill_value_from, skill_value_to = skill_value_to, skill_value_from
+                        skill_value_from, skill_value_to = (
+                            skill_value_to,
+                            skill_value_from,
+                        )
                         ppg_ratio_edge = 1 / ppg_ratio_edge
 
                     correlation_data[(skill_value_from, skill_value_to)].append(
@@ -282,16 +325,33 @@ def propagate_skill_levels(propagation_sequence):
         # Save correlation data to the database
         for (skill_value_from, skill_value_to), ppg_ratios in correlation_data.items():
             ppg_ratios = [(ppg_ratio, n_games) for ppg_ratio, n_games in ppg_ratios]
-            ppg_ratios_array = np.array(ppg_ratios, dtype=[('ppg_ratio', float), ('n_games', int)])
-            ppg_ratios_filtered = Config.discard_outliers(ppg_ratios_array['ppg_ratio'])
+            ppg_ratios_array = np.array(
+                ppg_ratios, dtype=[("ppg_ratio", float), ("n_games", int)]
+            )
+            ppg_ratios_filtered = Config.discard_outliers(ppg_ratios_array["ppg_ratio"])
             if len(ppg_ratios_filtered) > 0:
-                avg_ppg_ratio = float(sum(ppg_ratio * n_games for ppg_ratio, n_games in ppg_ratios if ppg_ratio in ppg_ratios_filtered) / sum(n_games for ppg_ratio, n_games in ppg_ratios if ppg_ratio in ppg_ratios_filtered))
-                total_n_games = sum(n_games for ppg_ratio, n_games in ppg_ratios if ppg_ratio in ppg_ratios_filtered)
+                avg_ppg_ratio = float(
+                    sum(
+                        ppg_ratio * n_games
+                        for ppg_ratio, n_games in ppg_ratios
+                        if ppg_ratio in ppg_ratios_filtered
+                    )
+                    / sum(
+                        n_games
+                        for ppg_ratio, n_games in ppg_ratios
+                        if ppg_ratio in ppg_ratios_filtered
+                    )
+                )
+                total_n_games = sum(
+                    n_games
+                    for ppg_ratio, n_games in ppg_ratios
+                    if ppg_ratio in ppg_ratios_filtered
+                )
                 correlation = SkillValuePPGRatio(
                     from_skill_value=skill_value_from,
                     to_skill_value=skill_value_to,
                     ppg_ratio=avg_ppg_ratio,
-                    n_games=total_n_games  # Store the sum of games
+                    n_games=total_n_games,  # Store the sum of games
                 )
                 session.add(correlation)
                 session.commit()
@@ -300,14 +360,22 @@ def propagate_skill_levels(propagation_sequence):
                 max_skill_value = max(max_skill_value, skill_value_from, skill_value_to)
 
     # Propagate skill levels
-    levels = session.query(Level).filter(Level.skill_propagation_sequence == propagation_sequence).all()
+    levels = (
+        session.query(Level)
+        .filter(Level.skill_propagation_sequence == propagation_sequence)
+        .all()
+    )
     suggested_skill_values = defaultdict(list)
 
     for level in levels:
-        edges = session.query(LevelsGraphEdge).filter(
-            (LevelsGraphEdge.from_level_id == level.id) |
-            (LevelsGraphEdge.to_level_id == level.id)
-        ).all()
+        edges = (
+            session.query(LevelsGraphEdge)
+            .filter(
+                (LevelsGraphEdge.from_level_id == level.id)
+                | (LevelsGraphEdge.to_level_id == level.id)
+            )
+            .all()
+        )
 
         for edge in edges:
             if edge.n_connections < Config.MIN_CONNECTIONS_FOR_PROPAGATION:
@@ -322,10 +390,14 @@ def propagate_skill_levels(propagation_sequence):
 
             target_level = session.query(Level).filter_by(id=target_level_id).first()
             if target_level and target_level.skill_propagation_sequence == -1:
-                correlations = session.query(SkillValuePPGRatio).filter(
-                    (SkillValuePPGRatio.from_skill_value <= level.skill_value) &
-                    (SkillValuePPGRatio.to_skill_value >= level.skill_value)
-                ).all()
+                correlations = (
+                    session.query(SkillValuePPGRatio)
+                    .filter(
+                        (SkillValuePPGRatio.from_skill_value <= level.skill_value)
+                        & (SkillValuePPGRatio.to_skill_value >= level.skill_value)
+                    )
+                    .all()
+                )
 
                 if correlations:
                     weighted_skill_values = []
@@ -334,17 +406,24 @@ def propagate_skill_levels(propagation_sequence):
                         # Let's avoid extrapolating from the end of the edge and away from the edge!
 
                         # Check left side of the edge
-                        if (level.skill_value == correlation.from_skill_value and level.skill_value > min_skill_value):
+                        if (
+                            level.skill_value == correlation.from_skill_value
+                            and level.skill_value > min_skill_value
+                        ):
                             if ppg_ratio_edge < 1:
                                 continue
                         # Check right side of the edge
-                        if (level.skill_value == correlation.to_skill_value and level.skill_value < max_skill_value):
+                        if (
+                            level.skill_value == correlation.to_skill_value
+                            and level.skill_value < max_skill_value
+                        ):
                             if ppg_ratio_edge > 1:
                                 continue
 
-                        
                         # First confirm which way are we going here
-                        if (ppg_ratio_edge < 1 and correlation.ppg_ratio > 1) or (ppg_ratio_edge > 1 and correlation.ppg_ratio < 1):
+                        if (ppg_ratio_edge < 1 and correlation.ppg_ratio > 1) or (
+                            ppg_ratio_edge > 1 and correlation.ppg_ratio < 1
+                        ):
                             # Reverse the correlation
                             from_skill_value = correlation.to_skill_value
                             to_skill_value = correlation.from_skill_value
@@ -360,12 +439,16 @@ def propagate_skill_levels(propagation_sequence):
                             ppg_ratio_range = 1 / ppg_ratio_range
                         else:
                             ppg_ratio_for_extrapolation = ppg_ratio_edge
-                        
+
                         # Interpolate or extrapolate skill value
                         skill_value_range = to_skill_value - from_skill_value
-                        skill_value_diff = (ppg_ratio_for_extrapolation / ppg_ratio_range) * skill_value_range
+                        skill_value_diff = (
+                            ppg_ratio_for_extrapolation / ppg_ratio_range
+                        ) * skill_value_range
                         new_skill_value = level.skill_value + skill_value_diff
-                        weighted_skill_values.append((new_skill_value, correlation.n_games))
+                        weighted_skill_values.append(
+                            (new_skill_value, correlation.n_games)
+                        )
                         # if target_level.id == 229:
                         #     print(f"Debug: From Level ID: {level.id}, To Level ID: {target_level.id}")
                         #     print(f"Debug: From Skill Value: {level.skill_value} PPG Ratio: {ppg_ratio_for_extrapolation}, PPG Ratio Range: {ppg_ratio_range}")
@@ -374,8 +457,16 @@ def propagate_skill_levels(propagation_sequence):
 
                     # Calculate weighted average of new skill values
                     total_n_games = sum(n_games for _, n_games in weighted_skill_values)
-                    weighted_avg_skill_value = sum(skill_value * n_games for skill_value, n_games in weighted_skill_values) / total_n_games
-                    suggested_skill_values[target_level_id].append(weighted_avg_skill_value)
+                    weighted_avg_skill_value = (
+                        sum(
+                            skill_value * n_games
+                            for skill_value, n_games in weighted_skill_values
+                        )
+                        / total_n_games
+                    )
+                    suggested_skill_values[target_level_id].append(
+                        weighted_avg_skill_value
+                    )
 
     # Update skill values for target levels
     session.flush()  # Ensure all previous changes are flushed before updates
@@ -387,10 +478,12 @@ def propagate_skill_levels(propagation_sequence):
             if avg_skill_value < min_skill_value:
                 avg_skill_value = min_skill_value - 0.01
             try:
-                session.query(Level).filter_by(id=target_level_id).update({
-                    'skill_value': avg_skill_value,
-                    'skill_propagation_sequence': propagation_sequence + 1
-                })
+                session.query(Level).filter_by(id=target_level_id).update(
+                    {
+                        "skill_value": avg_skill_value,
+                        "skill_propagation_sequence": propagation_sequence + 1,
+                    }
+                )
                 session.flush()  # Flush each update individually
             except Exception as e:
                 print(f"Error updating level {target_level_id}: {e}")
@@ -400,6 +493,7 @@ def propagate_skill_levels(propagation_sequence):
 
     print(f"Skill levels have been propagated for sequence {propagation_sequence}.")
 
+
 if __name__ == "__main__":
     reset_skill_values_in_divisions()
     build_levels_graph_edges()