"""量化投注引擎（EV、泊松預測、海拔修正）。"""

from __future__ import annotations

import math

import numpy as np
import pandas as pd
from scipy.stats import poisson


def calculate_value_bet(true_prob: float, decimal_odds: float, *, stake: float = 1.0) -> tuple[float, bool]:
    """計算期望值（EV）並判斷是否屬於 Value Bet。

    EV 計算：EV = (勝率 * 利潤) - (敗率 * 本金)
    其中利潤 = decimal_odds - 1。

    Returns
    -------
    ev_pct: float
        以本金為基底的 EV 百分比（EV / stake）。
    is_value_bet: bool
        當 EV > 0.03（3%）回傳 True。
    """

    prob = float(true_prob)
    odds = float(decimal_odds)
    if not 0 <= prob <= 1 or odds <= 1 or stake <= 0:
        return 0.0, False

    profit = odds - 1
    ev = prob * profit - (1 - prob) * stake
    ev_pct = ev / stake
    return round(ev_pct, 6), ev_pct > 0.03


class PoissonPredictor:
    """球員進球分佈預測器（2x2 進球建模）。"""

    def __init__(
        self,
        home_attack: float,
        home_defense: float,
        away_attack: float,
        away_defense: float,
        league_avg_goals: float,
    ) -> None:
        self.home_attack = float(home_attack)
        self.home_defense = float(home_defense)
        self.away_attack = float(away_attack)
        self.away_defense = float(away_defense)
        self.league_avg_goals = float(league_avg_goals)

        # 以攻守乘積估算 λ，並限制在合理範圍避免極端值發散。
        home_lambda = league_avg_goals * (self.home_attack / max(self.away_defense, 0.01))
        away_lambda = league_avg_goals * (self.away_attack / max(self.home_defense, 0.01))
        self.home_lambda = float(np.clip(home_lambda, 0.02, 6.5))
        self.away_lambda = float(np.clip(away_lambda, 0.02, 6.5))

    def predict_exact_score(self, home_goals: int, away_goals: int) -> float:
        """回傳指定波膽（home_goals, away_goals）發生機率。"""

        p_home = poisson.pmf(home_goals, self.home_lambda)
        p_away = poisson.pmf(away_goals, self.away_lambda)
        return float(p_home * p_away)

    def predict_over_under_prob(self, line: float = 2.5, max_goals: int = 10) -> tuple[float, float]:
        """回傳（under, over）機率。"""

        goals = pd.MultiIndex.from_product(
            [range(max_goals + 1), range(max_goals + 1)],
            names=['home', 'away'],
        ).to_frame(index=False)

        def joint_prob(r: pd.Series) -> float:
            return float(poisson.pmf(r['home'], self.home_lambda) * poisson.pmf(r['away'], self.away_lambda))

        probs = goals.apply(joint_prob, axis=1)
        total_goals = goals['home'] + goals['away']
        under = float(probs[total_goals <= line].sum())
        over = float(probs[total_goals > line].sum())
        return under, over


def adjust_away_defense_for_altitude(
    base_defense_rating: float,
    venue_altitude_meters: float,
    *,
    is_second_half: bool,
    penalty_factor: float = 0.35,
) -> float:
    """高海拔下修正客隊防守能力。

    當場地海拔高於 1500m 且處於下半場，套用對數懲罰，
    代表客隊在氧氣濃度降低下體能下降導致防守效率衰退。
    """

    base = float(base_defense_rating)
    if venue_altitude_meters <= 1500 or not is_second_half:
        return base

    # 以 log(1 + altitude/1000) 做平滑遞增函式，避免低海拔時劇烈改變。
    altitude_penalty = penalty_factor * math.log1p(venue_altitude_meters / 1000)
    return base * (1 - min(max(altitude_penalty, 0), 0.45))


__all__ = [
    'calculate_value_bet',
    'PoissonPredictor',
    'adjust_away_defense_for_altitude',
]