analytics/props_view_model.py

"""
analytics/props_view_model.py

Page-ready HR props view models used by the Props page redesign.

This module keeps grouping, sorting, and "featured props" selection out of the
render layer so the upcoming UI can consume stable structures instead of raw
rows.
"""

from __future__ import annotations

from typing import Any

import pandas as pd


def _coerce_int(value: Any, default: int = 0) -> int:
    try:
        if value is None or str(value).strip() in {"", "nan", "None"}:
            return default
        return int(float(value))
    except Exception:
        return default


def _coerce_bool(value: Any) -> bool:
    if isinstance(value, bool):
        return value
    text = str(value).strip().lower()
    return text in {"1", "true", "yes"}


def _make_game_key(row: pd.Series) -> str:
    event_id = str(row.get("event_id") or "").strip()
    if event_id and event_id not in {"nan", "None"}:
        return event_id
    away = str(row.get("away_team") or "").strip()
    home = str(row.get("home_team") or "").strip()
    commence = str(row.get("commence_time") or "").strip()
    return f"{away}|{home}|{commence}"


def _make_player_key(row: pd.Series) -> str:
    return f"{_make_game_key(row)}|{str(row.get('player_name') or '').strip().lower()}"


def _market_family_series(df: pd.DataFrame) -> pd.Series:
    if "market_family" in df.columns:
        return df["market_family"].astype(str).str.lower()
    if "market" in df.columns:
        return df["market"].astype(str).str.lower()
    return pd.Series([""] * len(df), index=df.index, dtype="object")


def _threshold_series(df: pd.DataFrame) -> pd.Series:
    if "threshold" in df.columns:
        source = df["threshold"]
    else:
        source = pd.Series([1] * len(df), index=df.index)
    return source.apply(lambda v: _coerce_int(v, default=1))


def _modeled_series(df: pd.DataFrame) -> pd.Series:
    if "is_modeled" in df.columns:
        source = df["is_modeled"]
    else:
        source = pd.Series([True] * len(df), index=df.index)
    return source.apply(_coerce_bool)


def _has_model_probability_series(df: pd.DataFrame) -> pd.Series:
    if "has_model_probability" in df.columns:
        source = df["has_model_probability"]
        return source.apply(_coerce_bool)
    if "model_hr_prob" in df.columns:
        return df["model_hr_prob"].notna()
    return pd.Series([False] * len(df), index=df.index)


def _modeled_hr_primary_series(df: pd.DataFrame) -> pd.Series:
    return (
        (_market_family_series(df) == "hr")
        & (_threshold_series(df) == 1)
        & _modeled_series(df)
    )


def _modeled_hr_primary_with_probability_series(df: pd.DataFrame) -> pd.Series:
    return _modeled_hr_primary_series(df) & _has_model_probability_series(df)


def _modeled_strikeout_with_probability_series(df: pd.DataFrame) -> pd.Series:
    market_series = _market_family_series(df)
    modeled_series = _modeled_series(df)
    if "has_model_probability" in df.columns:
        probability_series = df["has_model_probability"].apply(_coerce_bool)
    elif "fair_prob" in df.columns:
        probability_series = pd.to_numeric(df["fair_prob"], errors="coerce").notna()
    elif "model_k_prob" in df.columns:
        probability_series = pd.to_numeric(df["model_k_prob"], errors="coerce").notna()
    else:
        probability_series = pd.Series([False] * len(df), index=df.index)
    return (market_series == "k") & modeled_series & probability_series


def _sort_props_df(df: pd.DataFrame) -> pd.DataFrame:
    if df.empty:
        return df

    out = df.copy()
    if "threshold" not in out.columns:
        out["threshold"] = 1
    if "edge" not in out.columns:
        out["edge"] = None
    if "odds_american" not in out.columns:
        out["odds_american"] = None
    if "bet_ev" not in out.columns:
        out["bet_ev"] = None
    if "confidence_score" not in out.columns:
        out["confidence_score"] = None

    out["_threshold_sort"] = out["threshold"].apply(lambda v: _coerce_int(v, default=99))
    out["_ev_sort"] = pd.to_numeric(out["bet_ev"], errors="coerce").fillna(-999.0)
    out["_edge_sort"] = pd.to_numeric(out["edge"], errors="coerce").fillna(-999.0)
    out["_conf_sort"] = pd.to_numeric(out["confidence_score"], errors="coerce").fillna(-999.0)
    out["_odds_sort"] = pd.to_numeric(out["odds_american"], errors="coerce").fillna(-99999.0)
    out = out.sort_values(
        ["_threshold_sort", "_ev_sort", "_edge_sort", "_conf_sort", "_odds_sort"],
        ascending=[True, False, False, False, False],
        na_position="last",
    )
    return out.drop(columns=["_threshold_sort", "_ev_sort", "_edge_sort", "_conf_sort", "_odds_sort"])


def _compute_feature_score(df: pd.DataFrame) -> pd.DataFrame:
    if df.empty:
        return df

    out = df.copy()
    ev = pd.to_numeric(out.get("bet_ev"), errors="coerce").fillna(-9.0)
    edge = pd.to_numeric(out.get("edge"), errors="coerce").fillna(-9.0)
    confidence = pd.to_numeric(out.get("confidence_score"), errors="coerce").fillna(0.0)
    execution = pd.to_numeric(out.get("final_recommendation_score"), errors="coerce").fillna(0.0)

    out["featured_value_score"] = (
        (ev * 0.50)
        + (edge * 0.35)
        + (((confidence - 50.0) / 50.0) * 0.10)
        + (execution * 0.05)
    )
    return out


def _best_line_key(row: pd.Series) -> str:
    explicit_key = str(row.get("player_event_market_key") or "").strip()
    if explicit_key and explicit_key not in {"", "nan", "None"}:
        return explicit_key

    return "|".join(
        [
            str(row.get("event_id") or _make_game_key(row)).strip(),
            str(row.get("player_name") or "").strip().lower(),
            str(row.get("market_family") or row.get("market") or "").strip().lower(),
            str(_coerce_int(row.get("threshold"), default=1)),
        ]
    )


def select_best_lines_per_prop(mapped_df: pd.DataFrame) -> pd.DataFrame:
    """
    Collapse sportsbook rows down to the bettor-best line for each prop identity.

    Grouping stays threshold-aware, so 1+ HR and 2+ HR are never mixed.
    """
    if mapped_df is None or mapped_df.empty:
        return pd.DataFrame()

    working = mapped_df.copy()
    if "implied_prob" not in working.columns:
        working["implied_prob"] = None
    if "odds_american" not in working.columns:
        working["odds_american"] = None
    if "edge" not in working.columns:
        working["edge"] = None

    working["_best_line_key"] = working.apply(_best_line_key, axis=1)
    working["_implied_sort"] = pd.to_numeric(working["implied_prob"], errors="coerce").fillna(999.0)
    working["_edge_sort"] = pd.to_numeric(working["edge"], errors="coerce").fillna(-999.0)
    working["_odds_sort"] = pd.to_numeric(working["odds_american"], errors="coerce").fillna(-99999.0)

    working = working.sort_values(
        ["_best_line_key", "_implied_sort", "_edge_sort", "_odds_sort"],
        ascending=[True, True, False, False],
        na_position="last",
    )
    best = working.drop_duplicates(subset=["_best_line_key"], keep="first").copy()
    return best.drop(columns=["_best_line_key", "_implied_sort", "_edge_sort", "_odds_sort"]).reset_index(drop=True)


def build_featured_hr_props_df(mapped_df: pd.DataFrame, limit: int = 8) -> pd.DataFrame:
    if mapped_df is None or mapped_df.empty:
        return pd.DataFrame()

    featured = mapped_df.copy()
    featured = featured[_modeled_hr_primary_with_probability_series(featured)]

    if featured.empty:
        return pd.DataFrame()

    featured = select_best_lines_per_prop(featured)
    featured = _compute_feature_score(featured)

    sort_cols: list[str] = []
    ascending: list[bool] = []
    if "featured_value_score" in featured.columns:
        sort_cols.append("featured_value_score")
        ascending.append(False)
    if "bet_ev" in featured.columns:
        sort_cols.append("bet_ev")
        ascending.append(False)
    if "final_recommendation_score" in featured.columns:
        sort_cols.append("final_recommendation_score")
        ascending.append(False)
    if "edge" in featured.columns:
        sort_cols.append("edge")
        ascending.append(False)
    if "odds_american" in featured.columns:
        sort_cols.append("odds_american")
        ascending.append(False)
    if sort_cols:
        featured = featured.sort_values(sort_cols, ascending=ascending, na_position="last")

    return featured.head(max(1, int(limit))).reset_index(drop=True)


def build_best_on_slate_df(mapped_df: pd.DataFrame, limit: int = 8) -> pd.DataFrame:
    if mapped_df is None or mapped_df.empty:
        return pd.DataFrame()

    best_on_slate_full_df = _build_best_on_slate_full_df(mapped_df)
    if best_on_slate_full_df.empty:
        return pd.DataFrame()
    return best_on_slate_full_df.head(max(1, int(limit))).reset_index(drop=True)


def build_best_on_slate_summary(mapped_df: pd.DataFrame) -> dict[str, Any]:
    best_on_slate_full_df = _build_best_on_slate_full_df(mapped_df)
    if best_on_slate_full_df.empty:
        return {
            "modeled_props_count": 0,
            "sportsbooks_count": 0,
            "markets_count": 0,
            "best_ev": None,
            "best_edge": None,
        }

    return {
        "modeled_props_count": int(len(best_on_slate_full_df)),
        "sportsbooks_count": int(best_on_slate_full_df["sportsbook"].dropna().astype(str).nunique()) if "sportsbook" in best_on_slate_full_df.columns else 0,
        "markets_count": int(_market_family_series(best_on_slate_full_df).replace("", pd.NA).dropna().nunique()),
        "best_ev": pd.to_numeric(best_on_slate_full_df.get("bet_ev"), errors="coerce").dropna().max() if "bet_ev" in best_on_slate_full_df.columns else None,
        "best_edge": pd.to_numeric(best_on_slate_full_df.get("edge"), errors="coerce").dropna().max() if "edge" in best_on_slate_full_df.columns else None,
    }


def _build_best_on_slate_full_df(mapped_df: pd.DataFrame) -> pd.DataFrame:
    if mapped_df is None or mapped_df.empty:
        return pd.DataFrame()

    working = mapped_df.copy()
    eligible = working[
        _modeled_hr_primary_with_probability_series(working)
        | _modeled_strikeout_with_probability_series(working)
    ].copy()
    if eligible.empty:
        return pd.DataFrame()

    eligible = select_best_lines_per_prop(eligible)
    eligible = _compute_feature_score(eligible)

    sort_cols: list[str] = []
    ascending: list[bool] = []
    for col in ("bet_ev", "edge", "confidence_score", "final_recommendation_score", "featured_value_score"):
        if col in eligible.columns:
            sort_cols.append(col)
            ascending.append(False)
    if "odds_american" in eligible.columns:
        sort_cols.append("odds_american")
        ascending.append(False)
    if sort_cols:
        eligible = eligible.sort_values(sort_cols, ascending=ascending, na_position="last")

    return eligible.reset_index(drop=True)


def build_games_summary_df(mapped_df: pd.DataFrame) -> pd.DataFrame:
    if mapped_df is None or mapped_df.empty:
        return pd.DataFrame()

    working = mapped_df.copy()
    working["_game_key"] = working.apply(_make_game_key, axis=1)

    summary_rows: list[dict[str, Any]] = []
    for game_key, game_df in working.groupby("_game_key", dropna=False):
        primary_modeled = game_df[
            _modeled_hr_primary_with_probability_series(game_df)
        ].copy()
        primary_modeled = _sort_props_df(select_best_lines_per_prop(primary_modeled))
        if primary_modeled.empty:
            continue

        top_row = primary_modeled.iloc[0].to_dict() if not primary_modeled.empty else {}
        first_row = game_df.iloc[0]
        summary_rows.append(
            {
                "game_key": game_key,
                "event_id": first_row.get("event_id"),
                "away_team": first_row.get("away_team"),
                "home_team": first_row.get("home_team"),
                "commence_time": first_row.get("commence_time"),
                "modeled_props_count": int(len(primary_modeled)),
                "players_count": int(game_df["player_name"].nunique()),
                "best_edge": top_row.get("edge"),
                "best_bet_ev": top_row.get("bet_ev"),
                "top_confidence_score": top_row.get("confidence_score"),
                "top_player_name": top_row.get("player_name"),
                "top_display_label": top_row.get("display_label"),
                "top_book": top_row.get("sportsbook"),
                "top_odds_american": top_row.get("odds_american"),
                "top_verdict": top_row.get("verdict"),
                "top_market_family": top_row.get("market_family"),
            }
        )

    summary_df = pd.DataFrame(summary_rows)
    if summary_df.empty:
        return summary_df

    if "best_edge" in summary_df.columns:
        summary_df = summary_df.sort_values("best_edge", ascending=False, na_position="last")
    return summary_df.reset_index(drop=True)


def build_player_prop_detail_map(mapped_df: pd.DataFrame) -> dict[str, dict[str, Any]]:
    if mapped_df is None or mapped_df.empty:
        return {}

    working = _sort_props_df(mapped_df.copy())
    detail_map: dict[str, dict[str, Any]] = {}

    working["_player_key"] = working.apply(_make_player_key, axis=1)
    for player_key, player_df in working.groupby("_player_key", dropna=False):
        first_row = player_df.iloc[0]
        threshold_values = _threshold_series(player_df)
        primary_rows = player_df[threshold_values == 1].copy()
        alt_rows = player_df[threshold_values > 1].copy()
        modeled_primary_rows = primary_rows[_modeled_series(primary_rows)].copy()
        modeled_primary_rows = modeled_primary_rows[_has_model_probability_series(modeled_primary_rows)].copy()
        best_primary_rows = select_best_lines_per_prop(modeled_primary_rows if not modeled_primary_rows.empty else primary_rows)

        best_primary = best_primary_rows.iloc[0].to_dict() if not best_primary_rows.empty else None
        best_edge = best_primary.get("edge") if best_primary else None
        best_ev = best_primary.get("bet_ev") if best_primary else None
        best_book = best_primary.get("sportsbook") if best_primary else None
        best_odds = best_primary.get("odds_american") if best_primary else None

        detail_map[player_key] = {
            "player_key": player_key,
            "game_key": _make_game_key(first_row),
            "event_id": first_row.get("event_id"),
            "away_team": first_row.get("away_team"),
            "home_team": first_row.get("home_team"),
            "commence_time": first_row.get("commence_time"),
            "player_name": first_row.get("player_name"),
            "player_name_raw": first_row.get("player_name_raw"),
            "has_modeled_row": not modeled_primary_rows.empty,
            "has_alt_ladders": not alt_rows.empty,
            "best_edge": best_edge,
            "best_bet_ev": best_ev,
            "best_book": best_book,
            "best_odds_american": best_odds,
            "best_primary": best_primary,
            "best_primary_row": best_primary,
            "best_verdict": best_primary.get("verdict") if best_primary else None,
            "model_voice": best_primary.get("model_voice") if best_primary else None,
            "model_voice_primary_reason": best_primary.get("model_voice_primary_reason") if best_primary else None,
            "model_voice_caveat": best_primary.get("model_voice_caveat") if best_primary else None,
            "model_voice_for": best_primary.get("model_voice_for") if best_primary else None,
            "model_voice_against": best_primary.get("model_voice_against") if best_primary else None,
            "primary_rows": primary_rows.to_dict("records"),
            "alt_rows": alt_rows.to_dict("records"),
            "all_rows": player_df.to_dict("records"),
        }

    return detail_map


def _build_player_prop_detail_map_from_sorted(working: pd.DataFrame) -> dict[str, dict[str, Any]]:
    if working is None or working.empty:
        return {}

    detail_map: dict[str, dict[str, Any]] = {}
    if "_player_key" not in working.columns:
        working = working.copy()
        working["_player_key"] = working.apply(_make_player_key, axis=1)
    for player_key, player_df in working.groupby("_player_key", dropna=False):
        first_row = player_df.iloc[0]
        threshold_values = _threshold_series(player_df)
        primary_rows = player_df[threshold_values == 1].copy()
        alt_rows = player_df[threshold_values > 1].copy()
        modeled_primary_rows = primary_rows[_modeled_series(primary_rows)].copy()
        modeled_primary_rows = modeled_primary_rows[_has_model_probability_series(modeled_primary_rows)].copy()
        best_primary_rows = select_best_lines_per_prop(modeled_primary_rows if not modeled_primary_rows.empty else primary_rows)

        best_primary = best_primary_rows.iloc[0].to_dict() if not best_primary_rows.empty else None
        best_edge = best_primary.get("edge") if best_primary else None
        best_ev = best_primary.get("bet_ev") if best_primary else None
        best_book = best_primary.get("sportsbook") if best_primary else None
        best_odds = best_primary.get("odds_american") if best_primary else None

        detail_map[player_key] = {
            "player_key": player_key,
            "game_key": _make_game_key(first_row),
            "event_id": first_row.get("event_id"),
            "away_team": first_row.get("away_team"),
            "home_team": first_row.get("home_team"),
            "commence_time": first_row.get("commence_time"),
            "player_name": first_row.get("player_name"),
            "player_name_raw": first_row.get("player_name_raw"),
            "has_modeled_row": not modeled_primary_rows.empty,
            "has_alt_ladders": not alt_rows.empty,
            "best_edge": best_edge,
            "best_bet_ev": best_ev,
            "best_book": best_book,
            "best_odds_american": best_odds,
            "best_primary": best_primary,
            "best_primary_row": best_primary,
            "best_verdict": best_primary.get("verdict") if best_primary else None,
            "model_voice": best_primary.get("model_voice") if best_primary else None,
            "model_voice_primary_reason": best_primary.get("model_voice_primary_reason") if best_primary else None,
            "model_voice_caveat": best_primary.get("model_voice_caveat") if best_primary else None,
            "model_voice_for": best_primary.get("model_voice_for") if best_primary else None,
            "model_voice_against": best_primary.get("model_voice_against") if best_primary else None,
            "primary_rows": primary_rows.to_dict("records"),
            "alt_rows": alt_rows.to_dict("records"),
            "all_rows": player_df.to_dict("records"),
        }

    return detail_map


def build_game_player_props_map(mapped_df: pd.DataFrame) -> dict[str, dict[str, Any]]:
    if mapped_df is None or mapped_df.empty:
        return {}

    working = _sort_props_df(mapped_df.copy())
    working["_game_key"] = working.apply(_make_game_key, axis=1)
    detail_map = build_player_prop_detail_map(working)

    game_map: dict[str, dict[str, Any]] = {}
    for game_key, game_df in working.groupby("_game_key", dropna=False):
        first_row = game_df.iloc[0]
        player_entries: list[dict[str, Any]] = []

        for _, player_seed_row in game_df.drop_duplicates(subset=["player_name"]).iterrows():
            player_key = _make_player_key(player_seed_row)
            detail = detail_map.get(player_key)
            if detail is None:
                continue
            best_primary = detail.get("best_primary") or {}
            player_entries.append(
                {
                    "player_key": player_key,
                    "player_name": detail.get("player_name"),
                    "player_name_raw": detail.get("player_name_raw"),
                    "has_modeled_row": detail.get("has_modeled_row", False),
                    "has_alt_ladders": detail.get("has_alt_ladders", False),
                    "best_edge": best_primary.get("edge"),
                    "best_book": best_primary.get("sportsbook"),
                    "best_odds_american": best_primary.get("odds_american"),
                    "best_model_hr_prob": best_primary.get("model_hr_prob") if pd.notna(best_primary.get("model_hr_prob")) else best_primary.get("fair_prob"),
                    "best_display_label": best_primary.get("display_label"),
                    "best_bet_ev": best_primary.get("bet_ev"),
                    "best_confidence_score": best_primary.get("confidence_score"),
                    "best_verdict": best_primary.get("verdict"),
                    "model_voice": best_primary.get("model_voice"),
                    "model_voice_primary_reason": best_primary.get("model_voice_primary_reason"),
                    "model_voice_caveat": best_primary.get("model_voice_caveat"),
                    "model_voice_for": best_primary.get("model_voice_for"),
                    "model_voice_against": best_primary.get("model_voice_against"),
                    "details": detail,
                }
            )

        player_entries = sorted(
            player_entries,
            key=lambda item: (
                item.get("best_edge") is None,
                -(float(item.get("best_edge") or -999.0)),
                str(item.get("player_name") or ""),
            ),
        )

        primary_modeled = game_df[
            _modeled_hr_primary_with_probability_series(game_df)
        ].copy()
        primary_modeled = _sort_props_df(select_best_lines_per_prop(primary_modeled))
        if primary_modeled.empty:
            continue
        top_row = primary_modeled.iloc[0].to_dict() if not primary_modeled.empty else {}

        game_map[game_key] = {
            "game_key": game_key,
            "event_id": first_row.get("event_id"),
            "away_team": first_row.get("away_team"),
            "home_team": first_row.get("home_team"),
            "commence_time": first_row.get("commence_time"),
            "modeled_props_count": int(len(primary_modeled)),
            "players_count": int(game_df["player_name"].nunique()),
            "best_edge": top_row.get("edge"),
            "best_bet_ev": top_row.get("bet_ev"),
            "top_confidence_score": top_row.get("confidence_score"),
            "top_player_name": top_row.get("player_name"),
            "top_display_label": top_row.get("display_label"),
            "top_book": top_row.get("sportsbook"),
            "top_odds_american": top_row.get("odds_american"),
            "top_verdict": top_row.get("verdict"),
            "top_market_family": top_row.get("market_family"),
            "players": player_entries,
        }

    return game_map


def build_hr_props_view_model(mapped_df: pd.DataFrame, featured_limit: int = 8) -> dict[str, Any]:
    if mapped_df is None or mapped_df.empty:
        return {
            "featured_props_df": pd.DataFrame(),
            "best_on_slate_df": pd.DataFrame(),
            "best_on_slate_summary": {
                "modeled_props_count": 0,
                "sportsbooks_count": 0,
                "markets_count": 0,
                "best_ev": None,
                "best_edge": None,
            },
            "games_summary_df": pd.DataFrame(),
            "game_player_props_map": {},
            "player_prop_detail_map": {},
        }

    working = _sort_props_df(mapped_df.copy())
    working["_game_key"] = working.apply(_make_game_key, axis=1)
    working["_player_key"] = working.apply(_make_player_key, axis=1)

    featured = working[_modeled_hr_primary_with_probability_series(working)].copy()
    if featured.empty:
        featured_props_df = pd.DataFrame()
    else:
        featured_props_df = select_best_lines_per_prop(featured)
        featured_props_df = _compute_feature_score(featured_props_df)
        sort_cols: list[str] = []
        ascending: list[bool] = []
        if "featured_value_score" in featured_props_df.columns:
            sort_cols.append("featured_value_score")
            ascending.append(False)
        if "bet_ev" in featured_props_df.columns:
            sort_cols.append("bet_ev")
            ascending.append(False)
        if "final_recommendation_score" in featured_props_df.columns:
            sort_cols.append("final_recommendation_score")
            ascending.append(False)
        if "edge" in featured_props_df.columns:
            sort_cols.append("edge")
            ascending.append(False)
        if "odds_american" in featured_props_df.columns:
            sort_cols.append("odds_american")
            ascending.append(False)
        if sort_cols:
            featured_props_df = featured_props_df.sort_values(sort_cols, ascending=ascending, na_position="last")
        featured_props_df = featured_props_df.head(max(1, int(featured_limit))).reset_index(drop=True)

    eligible = working[
        _modeled_hr_primary_with_probability_series(working)
        | _modeled_strikeout_with_probability_series(working)
    ].copy()
    if eligible.empty:
        best_on_slate_full_df = pd.DataFrame()
        best_on_slate_df = pd.DataFrame()
    else:
        best_on_slate_full_df = select_best_lines_per_prop(eligible)
        best_on_slate_full_df = _compute_feature_score(best_on_slate_full_df)
        sort_cols = []
        ascending = []
        for col in ("bet_ev", "edge", "confidence_score", "final_recommendation_score", "featured_value_score"):
            if col in best_on_slate_full_df.columns:
                sort_cols.append(col)
                ascending.append(False)
        if "odds_american" in best_on_slate_full_df.columns:
            sort_cols.append("odds_american")
            ascending.append(False)
        if sort_cols:
            best_on_slate_full_df = best_on_slate_full_df.sort_values(sort_cols, ascending=ascending, na_position="last")
        best_on_slate_df = best_on_slate_full_df.head(max(1, int(featured_limit))).reset_index(drop=True)

    if best_on_slate_full_df.empty:
        best_on_slate_summary = {
            "modeled_props_count": 0,
            "sportsbooks_count": 0,
            "markets_count": 0,
            "best_ev": None,
            "best_edge": None,
        }
    else:
        best_on_slate_summary = {
            "modeled_props_count": int(len(best_on_slate_full_df)),
            "sportsbooks_count": int(best_on_slate_full_df["sportsbook"].dropna().astype(str).nunique()) if "sportsbook" in best_on_slate_full_df.columns else 0,
            "markets_count": int(_market_family_series(best_on_slate_full_df).replace("", pd.NA).dropna().nunique()),
            "best_ev": pd.to_numeric(best_on_slate_full_df.get("bet_ev"), errors="coerce").dropna().max() if "bet_ev" in best_on_slate_full_df.columns else None,
            "best_edge": pd.to_numeric(best_on_slate_full_df.get("edge"), errors="coerce").dropna().max() if "edge" in best_on_slate_full_df.columns else None,
        }

    player_prop_detail_map = _build_player_prop_detail_map_from_sorted(working)

    summary_rows: list[dict[str, Any]] = []
    game_player_props_map: dict[str, dict[str, Any]] = {}
    for game_key, game_df in working.groupby("_game_key", dropna=False):
        primary_modeled = game_df[_modeled_hr_primary_with_probability_series(game_df)].copy()
        primary_modeled = _sort_props_df(select_best_lines_per_prop(primary_modeled))
        if primary_modeled.empty:
            continue
        top_row = primary_modeled.iloc[0].to_dict() if not primary_modeled.empty else {}
        first_row = game_df.iloc[0]

        summary_rows.append(
            {
                "game_key": game_key,
                "event_id": first_row.get("event_id"),
                "away_team": first_row.get("away_team"),
                "home_team": first_row.get("home_team"),
                "commence_time": first_row.get("commence_time"),
                "modeled_props_count": int(len(primary_modeled)),
                "players_count": int(game_df["player_name"].nunique()),
                "best_edge": top_row.get("edge"),
                "best_bet_ev": top_row.get("bet_ev"),
                "top_confidence_score": top_row.get("confidence_score"),
                "top_player_name": top_row.get("player_name"),
                "top_display_label": top_row.get("display_label"),
                "top_book": top_row.get("sportsbook"),
                "top_odds_american": top_row.get("odds_american"),
                "top_verdict": top_row.get("verdict"),
                "top_market_family": top_row.get("market_family"),
            }
        )

        player_entries: list[dict[str, Any]] = []
        for _, player_seed_row in game_df.drop_duplicates(subset=["player_name"]).iterrows():
            player_key = player_seed_row.get("_player_key") or _make_player_key(player_seed_row)
            detail = player_prop_detail_map.get(player_key)
            if detail is None:
                continue
            best_primary = detail.get("best_primary") or {}
            player_entries.append(
                {
                    "player_key": player_key,
                    "player_name": detail.get("player_name"),
                    "player_name_raw": detail.get("player_name_raw"),
                    "has_modeled_row": detail.get("has_modeled_row", False),
                    "has_alt_ladders": detail.get("has_alt_ladders", False),
                    "best_edge": best_primary.get("edge"),
                    "best_book": best_primary.get("sportsbook"),
                    "best_odds_american": best_primary.get("odds_american"),
                    "best_model_hr_prob": best_primary.get("model_hr_prob") if pd.notna(best_primary.get("model_hr_prob")) else best_primary.get("fair_prob"),
                    "best_display_label": best_primary.get("display_label"),
                    "best_bet_ev": best_primary.get("bet_ev"),
                    "best_confidence_score": best_primary.get("confidence_score"),
                    "best_verdict": best_primary.get("verdict"),
                    "model_voice": best_primary.get("model_voice"),
                    "model_voice_primary_reason": best_primary.get("model_voice_primary_reason"),
                    "model_voice_caveat": best_primary.get("model_voice_caveat"),
                    "model_voice_for": best_primary.get("model_voice_for"),
                    "model_voice_against": best_primary.get("model_voice_against"),
                    "details": detail,
                }
            )

        player_entries = sorted(
            player_entries,
            key=lambda item: (
                item.get("best_edge") is None,
                -(float(item.get("best_edge") or -999.0)),
                str(item.get("player_name") or ""),
            ),
        )

        game_player_props_map[game_key] = {
            "game_key": game_key,
            "event_id": first_row.get("event_id"),
            "away_team": first_row.get("away_team"),
            "home_team": first_row.get("home_team"),
            "commence_time": first_row.get("commence_time"),
            "modeled_props_count": int(len(primary_modeled)),
            "players_count": int(game_df["player_name"].nunique()),
            "best_edge": top_row.get("edge"),
            "best_bet_ev": top_row.get("bet_ev"),
            "top_confidence_score": top_row.get("confidence_score"),
            "top_player_name": top_row.get("player_name"),
            "top_display_label": top_row.get("display_label"),
            "top_book": top_row.get("sportsbook"),
            "top_odds_american": top_row.get("odds_american"),
            "top_verdict": top_row.get("verdict"),
            "top_market_family": top_row.get("market_family"),
            "players": player_entries,
        }

    games_summary_df = pd.DataFrame(summary_rows)
    if not games_summary_df.empty and "best_edge" in games_summary_df.columns:
        games_summary_df = games_summary_df.sort_values("best_edge", ascending=False, na_position="last").reset_index(drop=True)

    return {
        "featured_props_df": featured_props_df,
        "best_on_slate_df": best_on_slate_df,
        "best_on_slate_summary": best_on_slate_summary,
        "games_summary_df": games_summary_df,
        "game_player_props_map": game_player_props_map,
        "player_prop_detail_map": player_prop_detail_map,
    }