import pandas as pd from typing import Optional def calculate_luck(df: pd.DataFrame, as_of_week: Optional[int] = None, include_current: bool = False) -> pd.DataFrame: if as_of_week: df_complete = df[df.week <= as_of_week] else: status_list = ["postevent"] if include_current: status_list.append("midevent") df_complete = df[df.matchup_status.isin(status_list)] df_complete["actual_wins"] = df_complete["win_probability"].apply(lambda x: x > 0.5) df_list = [] n_teams = df.team_name.nunique() for week, df_week in df_complete.groupby("week"): if len(df_week) != n_teams: next else: df_week["against_all_wins"] = ((df_week.team_points.rank().astype("float") - 1) / (n_teams - 1)).round(2) df_week["against_all_losses"] = 1 - df_week["against_all_wins"] df_week["half_wins"] = (df_week["against_all_wins"] >= 0.5) * 1.0 df_week["half_losses"] = 1 - df_week["half_wins"] df_week["against_all_luck"] = df_week["actual_wins"] - df_week["against_all_wins"] df_week["half_luck"] = df_week["actual_wins"] - df_week["half_wins"] df_week["earned_wins"] = ((df_week["against_all_wins"] + df_week["half_wins"]) / 2).round(2) df_week["luck_wins"] = df_week["actual_wins"] - df_week["earned_wins"] df_list.append(df_week) df_luck = pd.concat(df_list) return df_luck def get_grouped_luck(df_luck_all_weeks: pd.DataFrame) -> pd.DataFrame: summ_cols = [ "team_name", "team_points", "against_all_wins", "half_wins", "actual_wins", "earned_wins", "luck_wins", ] sort_by = "luck_wins" return df_luck_all_weeks[summ_cols].groupby("team_name").sum().sort_values(sort_by, ascending=False) def summarize_remaining_wins_from_matches_map(matches_map): """ Return map for all teams to map of number remaining wins to array of playoff and bye prob, respectively. """ remaining_map = {} for team_name, team_matches_map in matches_map.items(): team_remaining_map = {} for match_binary_str, prob_list in team_matches_map.items(): n_wins = sum([int(x) for x in match_binary_str]) if n_wins in team_remaining_map: incr_obs, incr_playoff_prob, incr_bye_prob = prob_list if incr_obs == 0: continue current_obs, current_playoff_prob, current_bye_prob = team_remaining_map[n_wins] new_obs = current_obs + incr_obs new_playoff_prob = round( (current_obs * current_playoff_prob + incr_obs * incr_playoff_prob) / new_obs, 3 ) new_bye_prob = round((current_obs * current_bye_prob + incr_obs * incr_bye_prob) / new_obs, 3) team_remaining_map[n_wins] = [new_obs, new_playoff_prob, new_bye_prob] else: team_remaining_map[n_wins] = prob_list remaining_map[team_name] = team_remaining_map return remaining_map