Implement stat mapping

src/queries/nflverse/github_data.py

@@ -89,10 +89,29 @@ def get_player_kicking_stats(season_int: int | str = SEASON) -> pd.DataFrame:
     return df
 
 
+def get_team_defense_stats(season_int: int | str = SEASON) -> pd.DataFrame:
+    df = duckdb.sql(
+        f"""
+        select
+            week
+            , team
+            , sum(def_sacks) as def_sacks
+            , sum(def_interceptions) as def_interceptions
+            , sum(def_tds) as def_tds
+            , sum(def_fumble_recovery_opp) as def_fumble_recovery_opp
+            , sum(def_safety) as def_safety
+        from player_stats_player_stats_def_{season_int}
+        group by week, team
+        """
+    ).df()
+    return df
+
+
 NFLVERSE_ASSETS = [
     ("weekly_rosters", f"roster_weekly_{SEASON}.parquet"),
     ("player_stats", f"player_stats_{SEASON}.parquet"),
     ("player_stats", f"player_stats_kicking_{SEASON}.parquet"),
+    ("player_stats", f"player_stats_def_{SEASON}.parquet"),
 ]
 
 

src/stats.py

@@ -0,0 +1,153 @@
+from dataclasses import dataclass
+import pandas as pd
+
+from domain.playoffs import PLAYOFF_TEAM_DEF_PLAYER
+from queries.nflverse.github_data import get_player_kicking_stats, get_player_stats, get_team_defense_stats
+
+
+@dataclass
+class StatType:
+    key: str
+    score: float
+
+
+RUSH_TD = StatType(key="RUSH TD", score=6.0)
+REC_TD = StatType(key="REC TD", score=6.0)
+OFF_FUM_TD = StatType(key="OFF FUM TD", score=6.0)
+PASS_TD = StatType(key="PASS TD", score=4.0)
+FG_0_49 = StatType(key="FG 0-49", score=3.0)
+FG_50_ = StatType(key="FG 50+", score=5.0)
+TWO_PT = StatType(key="2 PT", score=2.0)
+RECEPTION = StatType(key="REC", score=1.0)
+RUSH_YD = StatType(key="RUSH YD", score=0.1)
+REC_YD = StatType(key="REC YD", score=0.1)
+PASS_YD = StatType(key="PASS YD", score=0.04)
+XP = StatType(key="XP", score=1.0)
+FUM_LOST = StatType(key="FUM LOST", score=-2.0)
+PASS_INT = StatType(key="PASS INT", score=-2.0)
+RET_TD = StatType(key="RET TD", score=6.0)
+DEF_TD = StatType(key="DEF TD", score=6.0)
+DEF_INT = StatType(key="DEF INT", score=2.0)
+FUM_REC = StatType(key="FUM REC", score=2.0)
+SAFETY = StatType(key="SAFETY", score=2.0)
+SACK = StatType(key="SACK", score=1.0)
+PTS_ALLOW_0 = StatType(key="PTS ALLOW 0", score=10.0)
+PTS_ALLOW_1_6 = StatType(key="PTS ALLOW 1-6", score=7.0)
+PTS_ALLOW_7_13 = StatType(key="PTS ALLOW 7-13", score=4.0)
+PTS_ALLOW_14_20 = StatType(key="PTS ALLOW 14-20", score=1.0)
+PTS_ALLOW_21_27 = StatType(key="PTS ALLOW 21-27", score=0.0)
+PTS_ALLOW_28_34 = StatType(key="PTS ALLOW 28-34", score=-1.0)
+PTS_ALLOW_35_ = StatType(key="PTS ALLOW 35+", score=-4.0)
+TEAM_WIN = StatType(key="TEAM WIN", score=5.0)
+ST_TD = StatType(key="ST TD", score=6.0)
+
+
+NFLVERSE_STAT_COL_TO_ID: dict[str, str] = {
+    "passing_tds": PASS_TD.key,
+    "pasing_yards": PASS_YD.key,
+    "passing_2pt_conversions": TWO_PT.key,
+    "sack_fumbles_lost": FUM_LOST.key,
+    "interceptions": PASS_INT.key,
+    "rushing_tds": RUSH_TD.key,
+    "rushing_yards": RUSH_YD.key,
+    "rushing_2pt_conversions": TWO_PT.key,
+    "rushing_fumbles_lost": FUM_LOST.key,
+    "receptions": RECEPTION.key,
+    "receiving_tds": REC_TD.key,
+    "receiving_yards": REC_YD.key,
+    "receiving_2pt_conversions": TWO_PT.key,
+    "receiving_fumbles_lost": FUM_LOST.key,
+    "special_teams_tds": ST_TD.key,
+    "pat_made": XP.key,
+    "fg_made_0_19": FG_0_49.key,
+    "fg_made_20_29": FG_0_49.key,
+    "fg_made_30_39": FG_0_49.key,
+    "fg_made_40_49": FG_0_49.key,
+    "fg_made_50_59": FG_50_.key,
+    "fg_made_60_": FG_50_.key,
+    "def_sacks": SACK.key,
+    "def_interceptions": DEF_INT.key,
+    "def_tds": DEF_TD.key,
+    "def_fumble_recovery_opp": FUM_REC.key,
+    "def_safety": SAFETY.key,
+}
+
+NFLVERSE_STAT_WEEK_TO_PLAYOFF_WEEK = {
+    18: 1,
+    20: 2,
+    21: 3,
+    22: 4,
+}
+
+
+def add_stats_from_player_df_to_stat_map(df: pd.DataFrame, stat_map):
+    df_playoffs = df[df.week.isin(NFLVERSE_STAT_WEEK_TO_PLAYOFF_WEEK.keys())]
+    df_playoffs.week = df_playoffs.week.apply(lambda x: NFLVERSE_STAT_WEEK_TO_PLAYOFF_WEEK[x])
+    for week_player_id_tuple, row in df_playoffs.set_index(["week", "player_id"]).iterrows():
+        if isinstance(week_player_id_tuple, tuple):
+            week, player_id = week_player_id_tuple
+        else:
+            # this won't happen but makes mypy happy
+            continue
+        player_stats: dict[str, float] = {}
+        for k, v in row.to_dict().items():
+            if k in NFLVERSE_STAT_COL_TO_ID:
+                if (mapped_k := NFLVERSE_STAT_COL_TO_ID[k]) in player_stats:
+                    player_stats[mapped_k] += v
+                else:
+                    player_stats[mapped_k] = v
+
+        if player_id not in stat_map[week]:
+            stat_map[week][player_id] = player_stats
+        else:
+            stat_map[week][player_id].update(player_stats)
+
+
+def add_stats_from_team_def_df_to_stat_map(df: pd.DataFrame, stat_map):
+    short_team_names_to_player_id = {t.rosters_short_name: p for t, p in PLAYOFF_TEAM_DEF_PLAYER}
+    df_playoffs = df[
+        (df.week.isin(NFLVERSE_STAT_WEEK_TO_PLAYOFF_WEEK.keys()) & df.team.isin(short_team_names_to_player_id.keys()))
+    ]
+    df_playoffs.week = df_playoffs.week.apply(lambda x: NFLVERSE_STAT_WEEK_TO_PLAYOFF_WEEK[x])
+
+    for week_team_tuple, row in df_playoffs.set_index(["week", "team"]).iterrows():
+        if isinstance(week_team_tuple, tuple):
+            week, team = week_team_tuple
+        else:
+            # this won't happen but makes mypy happy
+            continue
+        player_stats: dict[str, float] = {}
+        player_id = short_team_names_to_player_id[team]
+        for k, v in row.to_dict().items():
+            if k in NFLVERSE_STAT_COL_TO_ID:
+                if (mapped_k := NFLVERSE_STAT_COL_TO_ID[k]) in player_stats:
+                    player_stats[mapped_k] += v
+                else:
+                    player_stats[mapped_k] = v
+
+        if player_id not in stat_map[week]:
+            stat_map[week][player_id] = player_stats
+        else:
+            stat_map[week][player_id].update(player_stats)
+
+
+def assemble_nflverse_stats() -> dict[int, dict[str, dict[str, float]]]:
+    # map week -> player_id -> stat_key -> stat value
+    stat_map: dict[int, dict[str, dict[str, float]]] = {w: {} for w in NFLVERSE_STAT_WEEK_TO_PLAYOFF_WEEK.values()}
+
+    df_player_stats = get_player_stats()
+    df_kicking_stats = get_player_kicking_stats()
+    df_def_stats = get_team_defense_stats()
+
+    add_stats_from_player_df_to_stat_map(df_player_stats, stat_map)
+    add_stats_from_player_df_to_stat_map(df_kicking_stats, stat_map)
+    add_stats_from_team_def_df_to_stat_map(df_def_stats, stat_map)
+
+    return stat_map
+
+
+if __name__ == "__main__":
+    from queries.nflverse.github_data import load_assets
+
+    load_assets()
+    assemble_nflverse_stats()