def test_calculate_team_ratings_increases_home_rating_and_decreases_away_rating_for_a_draw_where_away_is_rated_higher(
):
[home, away] = elo.calculate_team_ratings(k_factor=25,
home_team=1667.03,
away_team=1723.12,
home_goals=1,
away_goals=1)
assert home == 1669.03
assert away == 1721.12
def test_calculate_team_ratings_returns_new_higher_home_rating_and_lower_away_ratings_for_home_win(
):
[home, away] = elo.calculate_team_ratings(k_factor=25,
home_team=1567.03,
away_team=1463.12,
home_goals=2,
away_goals=1)
assert home == 1575.9
assert away == 1454.25
def __apply_historic_elo_ratings(df, historic_elos, soft_reset_factor,
goal_points: int) -> pd.DataFrame:
"""
Calculate rolling ELO ratings for teams based on results provided in data frame
"""
# Initialise a dictionary with default elos for each team
for team in df['homeTeam'].unique():
if team not in historic_elos.keys():
historic_elos[team] = 1500
current_elos = historic_elos.copy()
current_home_attack_elos = historic_elos.copy()
current_home_defence_elos = historic_elos.copy()
current_away_attack_elos = historic_elos.copy()
current_away_defence_elos = historic_elos.copy()
match_elos = {}
match_home_strength = {}
match_away_strength = {}
last_season = 0
# Loop over the rows in the DataFrame
for index, row in df.iterrows():
# Get the current year
current_season = row['season']
# If it is a new season, soft-reset elos
if current_season != last_season:
current_elos = __revert_elos_to_mean(current_elos,
soft_reset_factor)
current_home_attack_elos = __revert_elos_to_mean(
current_home_attack_elos, soft_reset_factor)
current_home_defence_elos = __revert_elos_to_mean(
current_home_defence_elos, soft_reset_factor)
current_away_attack_elos = __revert_elos_to_mean(
current_away_attack_elos, soft_reset_factor)
current_away_defence_elos = __revert_elos_to_mean(
current_away_defence_elos, soft_reset_factor)
# Get the Game ID
match_id = row['fixtureID']
# Get the team and opposition
home_team = row['homeTeam']
away_team = row['awayTeam']
# Get the team and opposition elo score
home_team_elo = current_elos[home_team]
away_team_elo = current_elos[away_team]
home_attack_elo = current_home_attack_elos[home_team]
home_defence_elo = current_home_defence_elos[home_team]
away_attack_elo = current_away_attack_elos[away_team]
away_defence_elo = current_away_defence_elos[away_team]
# Add the elos and probabilities our elos dictionary and elo_probs dictionary based on the Match ID
match_elos[match_id] = [home_team_elo, away_team_elo]
match_home_strength[match_id] = [home_attack_elo, home_defence_elo]
match_away_strength[match_id] = [away_attack_elo, away_defence_elo]
# Calculate the new elos of each team
new_home_team_elo, new_away_team_elo = elo.calculate_team_ratings(
25, home_team_elo, away_team_elo, row['homeGoals'],
row['awayGoals'])
new_home_attack, new_away_defence = elo.calculate_attack_and_defence_ratings(
goal_points, home_attack_elo, away_defence_elo, row['homeGoals'])
new_away_attack, new_home_defence = elo.calculate_attack_and_defence_ratings(
goal_points, away_attack_elo, home_defence_elo, row['awayGoals'])
# Update elos in elo dictionary
current_elos[home_team] = new_home_team_elo
current_elos[away_team] = new_away_team_elo
current_home_attack_elos[home_team] = new_home_attack
current_away_attack_elos[away_team] = new_away_attack
current_home_defence_elos[home_team] = new_home_defence
current_away_defence_elos[away_team] = new_away_defence
last_season = current_season
updated = (df.assign(
homeElo=lambda frame: frame.fixtureID.map(match_elos).str[0],
awayElo=lambda frame: frame.fixtureID.map(match_elos).str[1],
homeAttackStrength=lambda frame: frame.fixtureID.map(
match_home_strength).str[0],
homeDefenceStrength=lambda frame: frame.fixtureID.map(
match_home_strength).str[1],
awayAttackStrength=lambda frame: frame.fixtureID.map(
match_away_strength).str[0],
awayDefenceStrength=lambda frame: frame.fixtureID.map(
match_away_strength).str[1],
))
return updated