"""第3段階（期待値予想）: ベース予想の勝率に単勝オッズを掛けて期待値を算出する。

◎〇▲△の決定には使わない（堅実順位はベース予想で確定済み）。
"""
from __future__ import annotations

from ai.four_mark_ranking import apply_prob_odds_ranking, normalize_win_probabilities


def attach_win_odds(rows: list[dict], odds_map: dict[int, float]) -> int:
    attached = 0
    for row in rows:
        hn = row.get("horse_number")
        if hn is None:
            row["win_odds"] = None
            continue
        horse_number = int(hn)
        win_odds = odds_map.get(horse_number)
        if win_odds and win_odds > 0:
            row["win_odds"] = float(win_odds)
            attached += 1
        else:
            row["win_odds"] = None
    return attached


def apply_ev_forecast(
    rows: list[dict],
    *,
    enabled: bool = True,
    min_prob_for_value: float = 0.03,
    odds_cap: float = 80.0,
) -> None:
    """
    期待値予想: 各馬の win_rate × 単勝オッズ → expected_roi（%）。
    印・順位は変更しない。
    """
    for row in rows:
        row["expected_roi"] = None
        row["stage3_value_score"] = None
        row["expected_roi_aggressive"] = None
        prob = float(row.get("win_rate") or row.get("win_probability") or 0.0)
        if not enabled or row.get("win_odds") is None or prob <= 0:
            continue
        odds = min(float(row["win_odds"]), odds_cap)
        effective_prob = max(0.0, prob - min_prob_for_value)
        expected_roi = effective_prob * odds * 100.0
        row["expected_roi"] = round(expected_roi, 2)
        row["stage3_value_score"] = round(expected_roi - 100.0, 2)
        row["expected_roi_aggressive"] = row["expected_roi"]


def apply_stage3_value(
    rows: list[dict],
    enabled: bool = True,
    min_prob_for_value: float = 0.02,
    odds_cap: float = 80.0,
    aggressive_min_prob: float = 0.0,
    aggressive_odds_cap: float = 300.0,
    *,
    use_legacy_normalized_marks: bool = False,
    apply_prob_odds_marks: bool = False,
) -> None:
    """後方互換: 予想1〜5 等の旧パイプライン。"""
    if use_legacy_normalized_marks:
        normalize_win_probabilities(rows, score_key="score_accuracy")

    for row in rows:
        row["expected_roi"] = None
        row["stage3_value_score"] = None
        prob = float(row.get("win_probability") or row.get("stage3_prob") or row.get("win_rate") or 0.0)
        if enabled and row.get("win_odds") is not None:
            odds = min(float(row["win_odds"]), odds_cap)
            effective_prob = max(0.0, prob - min_prob_for_value)
            expected_roi = effective_prob * odds * 100.0
            row["expected_roi"] = round(expected_roi, 2)
            row["stage3_value_score"] = round(expected_roi - 100.0, 2)

    rows.sort(key=lambda x: float(x.get("score_accuracy", 0.0)), reverse=True)
    for idx, row in enumerate(rows, start=1):
        row["rank_accuracy"] = idx

    rows.sort(
        key=lambda x: (
            -999999.0 if x.get("expected_roi") is None else float(x["expected_roi"]),
            float(x.get("score_accuracy", 0.0)),
        ),
        reverse=True,
    )
    for idx, row in enumerate(rows, start=1):
        row["rank_value"] = idx

    for row in rows:
        row["expected_roi_aggressive"] = None
        row["rank_value_aggressive"] = None
        if enabled and row.get("win_odds") is not None:
            odds_a = min(float(row["win_odds"]), float(aggressive_odds_cap))
            prob = float(row.get("stage3_prob") or row.get("win_rate") or 0.0)
            eff_a = max(0.0, prob - float(aggressive_min_prob))
            row["expected_roi_aggressive"] = round(eff_a * odds_a * 100.0, 2)

    rows.sort(
        key=lambda x: (
            -999999.0 if x.get("expected_roi_aggressive") is None else float(x["expected_roi_aggressive"]),
            float(x.get("win_odds") or 0.0),
            float(x.get("score_accuracy", 0.0)),
        ),
        reverse=True,
    )
    for idx, row in enumerate(rows, start=1):
        row["rank_value_aggressive"] = idx

    if apply_prob_odds_marks:
        apply_prob_odds_ranking(rows)

    rows.sort(key=lambda x: int(x["rank_accuracy"]))
    for row in rows:
        row["rank"] = int(row["rank_accuracy"])
        row["score"] = float(row["score_accuracy"])