root

#3KFwz38F8yprofile / publicneue Version

von @glad-crow-1664 erstellt 21 days ago kein Ablauf 24.7 KB Syntax: python

Deine Änderungen erstellen eine neue Paste, die mit dieser verknüpft ist — das Original bleibt unverändert.

neue Version Deine Änderungen erstellen eine neue Paste, die mit dieser verknüpft ist — das Original bleibt unverändert.

Sichtbarkeit Im Nutzerfeed anzeigen Zugriff public

Änderungsnotiz (optional)

Abbrechen

# core.py
import numpy as np
from scipy.stats import poisson
from league_modules import module_registry, MODULE_EXECUTION_ORDER
from rule_engine import RuleEngine, Context

class FootballPredictor:
    def __init__(self, **kwargs):
        self.params = kwargs
        defaults = {
            'home_control_rate': 0.50,
            'away_press_rate': 0.25,
            'home_defense_loss_rate': 0.15,
            'away_counter_attacks': 2.0,
            'home_stadium_condition': 'normal',
            'is_old_team_return': False,
            'is_top_club': False,
            'var_controversy': False,
            'big_ball_rate_home': 0.50,
            'big_ball_rate_away': 0.50,
            'home_team': 'Home',
            'away_team': 'Away',
            'home_shots_avg': 0,
            'away_shots_avg': 0,
            'home_possession': 0.50,
            'away_possession': 0.50,
            'odds_home_handicap': 0,
            'odds_draw_handicap': 0,
            'odds_away_handicap': 0,
        }
        for key, val in defaults.items():
            if key not in self.params:
                self.params[key] = val

self.league_config = kwargs.get('league_config', {})
        self.modules_cfg = self.league_config.get('modules', {})
        self.top_clubs = self.league_config.get('top_clubs', [])

self.odds_h = self.params['odds_home']
        self.odds_d = self.params['odds_draw']
        self.odds_a = self.params['odds_away']
        self.round = self.params['season_round']

self.context = Context()
        self._init_context_from_config()
        self.result = {}

def _init_context_from_config(self):
        cfg = self.league_config.get('modules', {})
        self.context.set('rdi_threshold', cfg.get('rdi', {}).get('threshold', 0.30))
        self.context.set('uat_threshold', cfg.get('uat', {}).get('threshold', 0.75))
        self.context.set('obd_threshold', cfg.get('odds_deviation', {}).get('threshold', 0.15))
        self.context.set('ds_threshold', cfg.get('dynamic_pressure', {}).get('threshold', 420))
        self.context.set('acf_threshold', cfg.get('adversity', {}).get('threshold', 1.00))
        self.context.set('q_attack_boost', 1.0)
        self.context.set('fpa_boost', 1.0)
        self.context.set('kpa_boost', 1.0)
        self.context.set('tcm_boost', 1.0)
        self.context.set('dta_boost', 1.0)
        self.context.set('pwl_boost', 1.0)
        self.context.set('splitter_boost', 1.0)
        self.context.set('wfl_triggered', False)
        self.context.set('orc_triggered', False)
        self.context.set('btp_triggered', False)
        self.context.set('hfa_triggered', False)
        self.context.set('obd_triggered', False)
        self.context.set('drc_triggered', False)
        self.context.set('rdi_triggered', False)
        self.context.set('uat_triggered', False)
        self.context.set('score_corrections', {})

def _calc_match_probabilities(self, xg_h, xg_a, max_goals=6):
        ph = [poisson.pmf(k, xg_h) for k in range(max_goals+1)]
        pa = [poisson.pmf(k, xg_a) for k in range(max_goals+1)]
        score_probs = {}
        win_h = draw = win_a = 0.0
        for i in range(max_goals+1):
            for j in range(max_goals+1):
                p = ph[i] * pa[j]
                score_probs[(i,j)] = p
                if i > j:
                    win_h += p
                elif i == j:
                    draw += p
                else:
                    win_a += p
        total = win_h + draw + win_a
        if total > 0:
            win_h /= total
            draw /= total
            win_a /= total
        best_score = max(score_probs, key=score_probs.get)
        best_prob = score_probs[best_score]
        sorted_scores = sorted(score_probs.items(), key=lambda x: -x[1])
        return {
            'win_h': win_h, 'draw': draw, 'win_a': win_a,
            'best_score': best_score,
            'best_prob': best_prob,
            'score_probs': score_probs,
            'sorted_scores': sorted_scores
        }

def _generate_score_matrix(self, sorted_scores):
        core = sorted_scores[0][0] if sorted_scores else (0,0)
        defend = []
        for (i,j), p in sorted_scores:
            if i+j <= 2:
                defend.append((i,j))
                if len(defend) >= 3:
                    break
        if not defend:
            defend = [(0,0), (1,0), (1,1)]
        
        open_attack_cfg = self.league_config.get('open_attack', {})
        threshold = open_attack_cfg.get('threshold', 0.50)
        big_ball_h = self.params.get('big_ball_rate_home', 0.50)
        big_ball_a = self.params.get('big_ball_rate_away', 0.50)
        
        open_attack = (big_ball_h > threshold and big_ball_a > threshold)
        
        total_xg = self.result.get('xg_h_after_A', 0) + self.result.get('xg_a_after_A', 0)
        if not open_attack and total_xg > 2.8:
            open_attack = True
            self.result['_debug']['open_attack_forced'] = True
        
        attack = []
        if open_attack:
            for (i,j), p in sorted_scores:
                if i+j >= 3:
                    attack.append((i,j))
                    if len(attack) >= 3:
                        break
        
        return {'core': core, 'defend': defend, 'attack': attack, 'open_attack': open_attack}

def _generate_strategy(self):
        win_h = self.result['prob_win_h']
        draw = self.result['prob_draw']
        win_a = self.result['prob_win_a']
        strategy = {}
        if win_h > 0.5:
            strategy['Safe'] = f"Home Win ({self.odds_h:.2f})"
        elif draw > 0.4:
            strategy['Safe'] = f"Draw ({self.odds_d:.2f})"
        elif win_a > 0.5:
            strategy['Safe'] = f"Away Win ({self.odds_a:.2f})"
        else:
            strategy['Safe'] = "Double chance: Draw/Away"

if self.odds_h < 1.30:
            strategy['Balanced'] = "Asian Handicap -1 / -1.5"
        elif 1.90 <= self.odds_h <= 2.05 and self.result.get('ewc_trigger', False):
            strategy['Balanced'] = "Home Win + Under 2.5"
        elif 2.00 <= self.odds_h <= 2.30:
            strategy['Balanced'] = "Home Win or Draw"
        else:
            strategy['Balanced'] = "Home Win / Draw"

total_xg = self.result['xg_h_after_A'] + self.result['xg_a_after_A']
        if total_xg < 2.0:
            strategy['Goals'] = "1, 2"
        elif total_xg < 3.0:
            strategy['Goals'] = "2, 3"
        else:
            strategy['Goals'] = "3, 4"

half1 = self._calc_match_probabilities(self.result['xg_h_half1'], self.result['xg_a_half1'])
        half2 = self._calc_match_probabilities(self.result['xg_h_half2'], self.result['xg_a_half2'])
        h1_state = 'W' if half1['win_h'] > 0.4 else ('D' if half1['draw'] > 0.4 else 'L')
        h2_state = 'W' if half2['win_h'] > 0.4 else ('D' if half2['draw'] > 0.4 else 'L')
        strategy['HT/FT'] = f"{h1_state}/{h2_state}"
        self.result['strategy'] = strategy

def _generate_tags(self):
        tags = []
        if self.result.get('ewc_trigger', False): tags.append('Economic Win')
        if self.result.get('dsr_iron_draw', False): tags.append('Draw Alert')
        if self.odds_h < 1.30: tags.append('Deep Odds')
        if self.params['home_stadium_condition'] != 'normal': tags.append('Home Discount')
        if self.result.get('vtd_trap', False): tags.append('Trap Warning')
        if self.result.get('bpr_triggered', False): tags.append('Counter Risk')
        if self.result.get('open_attack', False): tags.append('Attack Mode')
        if self.context.get('hfa_triggered'): tags.append('HFA Boost')
        if self.context.get('btp_triggered'): tags.append('Clutch Alert')
        if self.context.get('obd_triggered'): tags.append('Odds Deviation')
        if self.context.get('orc_triggered'): tags.append('ORC Alert')
        if self.context.get('drc_triggered'): tags.append('DRC Alert')
        if self.context.get('wfl_triggered'): tags.append('WFL Signal')
        if not tags: tags.append('Regular')
        self.result['tags'] = tags

def _apply_injury_adjustment(self, xg_h, xg_a):
        home_injuries = self.params.get('home_injuries', 0)
        away_injuries = self.params.get('away_injuries', 0)
        home_key_missing = self.params.get('home_key_missing', 0)
        away_key_missing = self.params.get('away_key_missing', 0)

injury_cfg = self.league_config.get('injury', {})
        home_penalty = (home_injuries * injury_cfg.get('home_injury_penalty', 0.04) +
                        home_key_missing * injury_cfg.get('home_key_penalty', 0.08))
        away_penalty = (away_injuries * injury_cfg.get('away_injury_penalty', 0.04) +
                        away_key_missing * injury_cfg.get('away_key_penalty', 0.08))

min_xg = injury_cfg.get('min_xg_after_injury', 0.25)
        xg_h = max(min_xg, xg_h - home_penalty)
        xg_a = max(min_xg, xg_a - away_penalty)

if home_penalty > 0 or away_penalty > 0:
            self.result['_debug']['injury_penalty_h'] = round(home_penalty, 3)
            self.result['_debug']['injury_penalty_a'] = round(away_penalty, 3)
            self.result['_debug']['xg_h_before_injury'] = round(xg_h + home_penalty, 3)
            self.result['_debug']['xg_a_before_injury'] = round(xg_a + away_penalty, 3)

return xg_h, xg_a

def _apply_handicap_correction(self, score_probs, sorted_scores):
        handicap_line = self.params.get('handicap_line', 0)
        odds_home_h = self.params.get('odds_home_handicap', 1.90)
        odds_away_h = self.params.get('odds_away_handicap', 1.90)

if handicap_line == 0 or odds_home_h == 0 or odds_away_h == 0:
            return score_probs, sorted_scores

prob_home_cover = 1 / odds_home_h if odds_home_h > 0 else 0
        prob_away_cover = 1 / odds_away_h if odds_away_h > 0 else 0
        total = prob_home_cover + prob_away_cover
        if total > 0:
            prob_home_cover /= total
            prob_away_cover /= total

if handicap_line < 0:
            target_goals = abs(handicap_line)
            if prob_home_cover > 0.55:
                for (i, j), p in sorted_scores[:15]:
                    if i - j >= int(target_goals) + 1:
                        score_probs[(i, j)] *= 1.15
                self.result['handicap_signal'] = 'Home Cover (Big Win Expected)'
            else:
                for (i, j), p in sorted_scores[:15]:
                    if 1 <= i - j <= 2:
                        score_probs[(i, j)] *= 1.10
                self.result['handicap_signal'] = 'Home Win but Lose Cover (Small Win)'
        elif handicap_line > 0:
            if prob_away_cover > 0.55:
                for (i, j), p in sorted_scores[:15]:
                    if j - i >= int(handicap_line) + 1:
                        score_probs[(i, j)] *= 1.15
                self.result['handicap_signal'] = 'Away Cover (Big Win Expected)'
            else:
                for (i, j), p in sorted_scores[:15]:
                    if 1 <= j - i <= 2:
                        score_probs[(i, j)] *= 1.10
                self.result['handicap_signal'] = 'Away Win but Lose Cover (Small Win)'

new_sorted = sorted(score_probs.items(), key=lambda x: -x[1])
        return score_probs, new_sorted

# ==================== AB 线通用评分（修复版） ====================
    def _apply_ab_scoring(self, xg_h, xg_a):
        """
        通用 AB 线评分调度器（从配置读取权重，不写死任何联赛逻辑）
        基于 A 线（基本面）和 B 线（盘口）的交叉验证修正 xG
        """
        ab_cfg = self.league_config.get('ab_scoring', {})
        if not ab_cfg.get('active', False):
            return xg_h, xg_a

# 1. 计算 A 线（基本面）评分
        a_cfg = ab_cfg.get('a_line', {})
        home_power = self.params.get('home_power', 50)
        away_power = self.params.get('away_power', 50)
        
        # 基础战力值归一化 (50->0.5, 100->1.0)
        a_score_h = 0.5 + (home_power - 50) / 100
        a_score_a = 0.5 + (away_power - 50) / 100
        
        # 伤停修正（从配置读取惩罚系数）
        injuries_penalty = a_cfg.get('injuries_penalty', 0.05)
        home_inj = self.params.get('home_injuries', 0) * injuries_penalty
        away_inj = self.params.get('away_injuries', 0) * injuries_penalty
        a_score_h = max(0.1, a_score_h - home_inj)
        a_score_a = max(0.1, a_score_a - away_inj)
        
        # 战意修正（从配置读取）
        mot_boost = a_cfg.get('motivation_boost', 1.1)
        if self.params.get('motivation_home', 0.5) > 0.6:
            a_score_h = a_score_h * mot_boost
        if self.params.get('motivation_away', 0.5) > 0.6:
            a_score_a = a_score_a * mot_boost

# 2. 计算 B 线（盘口）评分
        b_cfg = ab_cfg.get('b_line', {})
        # 理论胜率（基于 ELO 或 power）
        elo_h = self.params.get('elo_home', 1500)
        elo_a = self.params.get('elo_away', 1500)
        if elo_h == 1500 and elo_a == 1500:
            # 如果没有 ELO，用 power 映射（50→1300, 100→1700）
            elo_h = 1300 + (home_power - 50) * 8
            elo_a = 1300 + (away_power - 50) * 8
        theo_prob_h = 1 / (1 + 10 ** ((elo_a - elo_h) / 400))
        theo_prob_a = 1 - theo_prob_h
        
        # 市场隐含概率（从赔率提取）
        implied_h = 1 / self.odds_h if self.odds_h > 0 else 0.5
        implied_a = 1 / self.odds_a if self.odds_a > 0 else 0.5
        
        # B 线评分 = 1 - 偏差（偏差越小，评分越高）
        b_score_h = 1 - min(1, abs(implied_h - theo_prob_h) * 2)
        b_score_a = 1 - min(1, abs(implied_a - theo_prob_a) * 2)

# 3. 交叉验证与置信度计算
        val_cfg = ab_cfg.get('validation', {})
        resonance_threshold = val_cfg.get('resonance_threshold', 0.1)
        divergence_threshold = val_cfg.get('divergence_threshold', 0.3)
        resonance_boost = val_cfg.get('resonance_boost', 1.2)
        divergence_penalty = val_cfg.get('divergence_penalty', 0.7)

# 计算 A/B 线对主胜和客胜评分的平均一致性
        diff_h = abs(a_score_h - b_score_h)
        diff_a = abs(a_score_a - b_score_a)
        avg_diff = (diff_h + diff_a) / 2

if avg_diff <= resonance_threshold:
            confidence_boost = resonance_boost
            self.result['ab_validation'] = 'resonance'
        elif avg_diff >= divergence_threshold:
            confidence_boost = divergence_penalty
            self.result['ab_validation'] = 'divergence'
        else:
            confidence_boost = 1.0
            self.result['ab_validation'] = 'neutral'

# 4. 将修正系数应用到 xG（修复版）
        if confidence_boost != 1.0:
            if confidence_boost > 1.0:
                # 共振：双方信心都提升
                if implied_h > theo_prob_h:
                    # 市场更看好主队 → 主队xG提升，客队适度提升
                    xg_h = xg_h * confidence_boost
                    xg_a = xg_a * (1 + (confidence_boost - 1) * 0.5)
                else:
                    # 市场更看好客队 → 客队xG提升，主队适度提升
                    xg_a = xg_a * confidence_boost
                    xg_h = xg_h * (1 + (confidence_boost - 1) * 0.5)
            else:
                # 背离：双方信心都降低（用同一个惩罚系数，而不是一方被放大）
                # 原来的 xg / confidence_boost（即 xg / 0.7）会放大 xG，这是 bug
                # 修正为双方都乘以 confidence_boost（即都降低），只是降低幅度不同
                if implied_h > theo_prob_h:
                    # 市场更看好主队，但模型与市场背离 → 主队xG降低，客队降得更多
                    xg_h = xg_h * confidence_boost
                    # 客队 × (confidence_boost ^ 1.5)，比主队降得更多
                    xg_a = xg_a * (confidence_boost * confidence_boost)
                else:
                    # 市场更看好客队，但模型与市场背离 → 客队xG降低，主队降得更多
                    xg_a = xg_a * confidence_boost
                    # 主队 × (confidence_boost ^ 1.5)，比客队降得更多
                    xg_h = xg_h * (confidence_boost * confidence_boost)

self.result['ab_confidence_boost'] = confidence_boost
            self.result['ab_scoring_active'] = True

return xg_h, xg_a

# ==================== 核心 predict 方法 ====================
    def predict(self):
        self.result['_debug'] = {}
        xg_h = self.params['base_xg_home']
        xg_a = self.params['base_xg_away']
        self.result['_debug']['xg_h_initial'] = round(xg_h, 3)
        self.result['_debug']['xg_a_initial'] = round(xg_a, 3)

# ----- 从配置读取 xG 调整 -----
        xg_adj = self.league_config.get('xg_adjustments', {})
        boost = xg_adj.get('default_boost', 1.10)
        boost_cfg = xg_adj.get('boost_conditions', {})
        odds_h = self.odds_h
        odds_a = self.odds_a
        league = self.params.get('league', '')

if odds_h < 1.40:
            boost += boost_cfg.get('deep_odds', 0.12)
        if odds_a > 4.00:
            boost += boost_cfg.get('big_away_odds', 0.10)
        if self.params.get('big_ball_rate_home', 0.5) > 0.55 and self.params.get('big_ball_rate_away', 0.5) > 0.55:
            boost += boost_cfg.get('both_attacking', 0.12)
        if self.params.get('relegation_zone', False):
            boost += boost_cfg.get('relegation_zone', 0.10)
        if self.params.get('first_leg_lead', 0) >= 2:
            boost += boost_cfg.get('second_leg_conservative', -0.20)
        if self.params.get('forward_missing', 0) >= 2:
            boost += boost_cfg.get('forward_missing_penalty', -0.08)

xg_h = xg_h * xg_adj.get('home_boost_factor', 1.0)
        xg_a = xg_a * xg_adj.get('away_penalty_factor', 1.0)
        if self.params.get('relegation_zone', False) and xg_adj.get('relegation_home_boost', 0):
            xg_h = xg_h + xg_adj.get('relegation_home_boost', 0)
            xg_h = max(xg_h, xg_adj.get('relegation_min_xg', 1.0))

boost = max(xg_adj.get('boost_min', 0.85), min(xg_adj.get('boost_max', 1.60), boost))
        xg_h = xg_h * boost
        xg_a = xg_a * boost
        self.result['_debug']['offensive_boost'] = round(boost, 3)
        self.result['_debug']['boost_reasons'] = []
        if odds_h < 1.40: self.result['_debug']['boost_reasons'].append('deep_odds')
        if odds_a > 4.00: self.result['_debug']['boost_reasons'].append('big_away_odds')
        if self.params.get('big_ball_rate_home', 0.5) > 0.55 and self.params.get('big_ball_rate_away', 0.5) > 0.55:
            self.result['_debug']['boost_reasons'].append('both_attacking')
        if self.params.get('first_leg_lead', 0) >= 2:
            self.result['_debug']['boost_reasons'].append('second_leg_conservative')

# 西甲深盘额外加成
        if league == 'LaLiga' and self.params.get('is_top_club', False) and odds_h < 1.40:
            la_liga_extra = xg_adj.get('la_liga_deep_odds_extra', 0.08)
            boost += la_liga_extra
            self.result['_debug']['la_liga_extra_boost'] = la_liga_extra

xg_max = xg_adj.get('xg_max', 5.0)
        xg_h = min(xg_h, xg_max)
        xg_a = min(xg_a, xg_max)

# 西甲豪门主场保底
        la_liga_floor = xg_adj.get('la_liga_home_floor', 2.2)
        if league == 'LaLiga' and self.params.get('is_top_club', False) and odds_h < 1.40:
            xg_h = max(xg_h, la_liga_floor)
            self.result['_debug']['la_liga_home_floor_applied'] = True
            self.result['_debug']['la_liga_home_floor_value'] = la_liga_floor

# 伤病调整
        xg_h, xg_a = self._apply_injury_adjustment(xg_h, xg_a)
        
        # ===== 🆕 AB 线评分（通用交叉验证） =====
        xg_h, xg_a = self._apply_ab_scoring(xg_h, xg_a)
        
        self.result['_debug']['xg_h_after_injury'] = round(xg_h, 3)
        self.result['_debug']['xg_a_after_injury'] = round(xg_a, 3)

# 通用主场保底（巴西杯等）
        if self.league_config.get('home_floor', {}).get('active', False):
            floor_cfg = self.league_config.get('home_floor', {})
            if self.params.get('is_top_club', False) and self.params.get('home_win_rate', 0) > floor_cfg.get('win_rate_threshold', 0.70):
                xg_h = max(xg_h, floor_cfg.get('min_xg', 1.5))
                self.result['_debug']['home_floor_applied'] = True

# ===== 模块执行（含沙特新模块） =====
        pre_modules = ['wfl', 'rdi', 'uat']
        for module_name in pre_modules:
            if module_name in module_registry:
                module_func = module_registry[module_name]
                xg_h, xg_a = module_func(self, xg_h, xg_a)
                self.context.set('current_xg_h', xg_h)
                self.context.set('current_xg_a', xg_a)
                RuleEngine.apply_rules(self.context)

for module_name in MODULE_EXECUTION_ORDER:
            if module_name in module_registry and module_name not in pre_modules:
                module_cfg = self.modules_cfg.get(module_name, {})
                if module_cfg.get('active', True):
                    module_func = module_registry[module_name]
                    #print(f"▶️ 执行模块: {module_name}")
                    xg_h, xg_a = module_func(self, xg_h, xg_a)
                    self.context.set('current_xg_h', xg_h)
                    self.context.set('current_xg_a', xg_a)
                    RuleEngine.apply_rules(self.context)
        # ===== 🆕 全局 xG 上限保护（防止极端强弱对话爆炸） =====
        xg_global_cap = xg_adj.get('xg_global_cap', 5.0)
        if xg_h > xg_global_cap:
            self.result['_debug']['xg_h_capped'] = round(xg_h, 3)
            xg_h = xg_global_cap
        if xg_a > xg_global_cap:
            self.result['_debug']['xg_a_capped'] = round(xg_a, 3)
            xg_a = xg_global_cap

self.result['xg_h_after_A'] = xg_h
        self.result['xg_a_after_A'] = xg_a
        self.result['_debug']['xg_h_final'] = round(xg_h, 3)
        self.result['_debug']['xg_a_final'] = round(xg_a, 3)

# 半全场划分
        tds_config = self.modules_cfg.get('tds', {})
        if tds_config.get('active', False):
            early = tds_config.get('early', [0.40, 0.60])
            mid = tds_config.get('mid', [0.45, 0.55])
            deep = tds_config.get('deep', [0.35, 0.65])
            if self.round <= 3:
                half_h, half_a = early[0], early[1]
            elif self.odds_h < 1.30:
                half_h, half_a = deep[0], deep[1]
            else:
                half_h, half_a = mid[0], mid[1]
        else:
            half_h, half_a = 0.45, 0.55

h_h1, h_h2 = xg_h * half_h, xg_h * half_a
        a_h1, a_h2 = xg_a * half_h, xg_a * half_a
        if self.params.get('var_controversy', False):
            a_h2 *= 1.05

self.result['xg_h_half1'] = h_h1
        self.result['xg_h_half2'] = h_h2
        self.result['xg_a_half1'] = a_h1
        self.result['xg_a_half2'] = a_h2

# 概率计算
        probs = self._calc_match_probabilities(xg_h, xg_a)
        self.result['core_prob'] = probs['best_prob']

score_corrections = self.context.get('score_corrections', {})
        for score_key, weight in score_corrections.items():
            if score_key in probs['score_probs']:
                probs['score_probs'][score_key] *= weight
        probs['sorted_scores'] = sorted(probs['score_probs'].items(), key=lambda x: -x[1])

win_h, draw, win_a = probs['win_h'], probs['draw'], probs['win_a']
        if self.result.get('dsr_iron_draw', False):
            draw *= 1.20
        if self.context.get('q_attack_boost', 1.0) != 1.0:
            win_h *= self.context.get('q_attack_boost', 1.0)
        if self.context.get('fpa_boost', 1.0) != 1.0:
            win_h *= self.context.get('fpa_boost', 1.0)

total = win_h + draw + win_a
        if total > 0:
            win_h /= total
            draw /= total
            win_a /= total

draw_cap = self.modules_cfg.get('dsr', {}).get('draw_cap', 0.35)
        if draw > draw_cap:
            draw = draw_cap
            total = win_h + win_a
            if total > 0:
                win_h = win_h / total * (1 - draw_cap)
                win_a = win_a / total * (1 - draw_cap)

self.result['prob_win_h'] = win_h
        self.result['prob_draw'] = draw
        self.result['prob_win_a'] = win_a

if self.params.get('handicap_line', 0) != 0:
            probs['score_probs'], probs['sorted_scores'] = self._apply_handicap_correction(
                probs['score_probs'], probs['sorted_scores']
            )

matrix = self._generate_score_matrix(probs['sorted_scores'])
        self.result['score_core'] = matrix['core']
        self.result['score_defend'] = matrix['defend']
        self.result['score_attack'] = matrix['attack']
        self.result['open_attack'] = matrix['open_attack']

self._generate_strategy()
        self._generate_tags()
        return self.result