def all_champ_all_lanes_win_rates(matches_df, file_name=''):
df = pd.DataFrame()
lanes = dc.get_lanes_roles()
for lane in lanes:
temp = all_champ_win_rates(matches_df, lane=lane)
df[lane + '_win_rate'] = temp['win_rate']
df[lane + '_games_played'] = temp['games_played']
if file_name != '':
df.to_csv(file_name)
return df
def all_champ_pairs_all_lanes(matches_df, file_name=''):
df = pd.DataFrame()
lanes = dc.get_lanes_roles()
for lane1 in lanes:
for lane2 in lanes:
if lane1 != lane2:
print(lane1 + '_' + lane2)
temp = all_champ_pairs_fixed_lane(matches_df, lane1, lane2)
df[lane1 + '_' + lane2 + '_win_rate'] = temp['win_rate']
df[lane1 + '_' + lane2 +
'_games_played'] = temp['games_played']
if file_name != '':
df.to_csv(file_name)
return df
def all_h2h_pairs_all_lanes(matches_df, file_name=''):
"""Produces all head to head win rates for all lane matchups -- even across different lanes
(eg. TOP_SOLO Renekton vs MID_SOLO Xerath)."""
df = pd.DataFrame()
lanes = dc.get_lanes_roles()
for lane1 in lanes:
print(lane1)
for lane2 in lanes:
print(lane1 + '_' + lane2)
temp = all_h2h_pairs_fixed_lane(matches_df, lane1, lane2)
df[lane1 + '_' + lane2 + '_win_rate'] = temp['win_rate']
df[lane1 + '_' + lane2 + '_games_played'] = temp['games_played']
if file_name != '':
df.to_csv(file_name)
return df
def all_champ_pairs_all_lanes(matches_df, file_name=''):
df = pd.DataFrame()
lanes = dc.get_lanes_roles()
for lane1 in lanes:
for lane2 in lanes:
if lane1 != lane2:
print(lane1 + '_' + lane2)
temp = all_champ_pairs_fixed_lane(matches_df, lane1, lane2)
df[lane1 + '_' + lane2 + '_wr'] = temp['win_rate']
df[lane1 + '_' + lane2 + '_gp'] = temp['games_played'].astype(int)
df[lane1 + '_' + lane2 + '_ra'] = temp['red_apps'].astype(int)
df[lane1 + '_' + lane2 + '_rw'] = temp['red_wins'].astype(int)
df[lane1 + '_' + lane2 + '_ba'] = temp['blue_apps'].astype(int)
df[lane1 + '_' + lane2 + '_bw'] = temp['blue_wins'].astype(int)
if file_name != '':
df.to_csv(file_name)
return df
def main():
# Load match data
data = pd.read_csv(
os.getenv('MINED_DATA_DIR') + 'processed_match_data.csv')
# Train, validation, test split
train = data.sample(int(.6 * data.shape[0]), random_state=410)
validation = data[~data['match_id'].isin(train['match_id'])].sample(
int(.2 * data.shape[0]), random_state=411)
test = data[~data['match_id'].isin(
pd.concat([train['match_id'], validation['match_id']],
axis=0,
ignore_index=True))]
# Build win rates (optional)
if len(sys.argv) > 1:
indiv_wr_boolean = sys.argv[1]
print('Fire 1')
if len(sys.argv) > 2:
paired_wr_boolean = sys.argv[2]
print('Fire 2')
if len(sys.argv) > 3:
h2h_wr_boolean = sys.argv[3]
print('Fire 3')
build_win_rates(train, indiv_wr_boolean, paired_wr_boolean, h2h_wr_boolean)
# Load win rates
indiv_win_rates = pd.read_csv(indiv_win_rate_file)
paired_win_rates = pd.read_csv(paired_win_rate_file)
h2h_win_rates = pd.read_csv(h2h_win_rate_file)
indiv_win_rates = indiv_win_rates.rename({'Unnamed: 0': 'champ'},
axis='columns')
paired_win_rates = paired_win_rates.rename({'Unnamed: 0': 'champ_pair'},
axis='columns')
h2h_win_rates = h2h_win_rates.rename({'Unnamed: 0': 'champ_pair'},
axis='columns')
#### Win rate significance ####
# Filter out low games played win rates with win_rates.win_rate_significance()
#### Join win rates ####
teams_lanes_roles = dc.get_teams_lanes_roles()
teams = dc.get_teams()
lanes_roles = dc.get_lanes_roles()
# Join individ win rates
for tlr in teams_lanes_roles:
# want to take index of indiv win rates and a single column of it
data = pd.merge(data,
indiv_win_rates[['champ', tlr[4:] + '_win_rate']],
how='left',
left_on=tlr,
right_on='champ')
data = data.rename({tlr[4:] + '_win_rate': tlr + '_wr'}, axis=1)
data = data.drop(['champ'], axis=1)
# Join paired win rates
teams = dc.get_teams()
for lr1 in lanes_roles:
print(lr1)
for lr2 in lanes_roles:
if lr1 != lr2:
for team in teams:
# create join key
tlr1 = team + '_' + lr1
tlr2 = team + '_' + lr2
data[tlr1 + '_' + tlr2] = data[tlr1].str.cat(data[tlr2],
sep='_')
data = pd.merge(data,
paired_win_rates[[
'champ_pair',
tlr1[4:] + '_' + tlr2[4:] + '_win_rate'
]],
how='left',
left_on=tlr1 + '_' + tlr2,
right_on='champ_pair')
data = data.rename(
{
tlr1[4:] + '_' + tlr2[4:] + '_win_rate':
tlr1 + '_' + tlr2[4:] + '_wr'
},
axis=1)
data = data.drop(['champ_pair', tlr1 + '_' + tlr2], axis=1)
# Join h2h win rates
for lr1 in lanes_roles:
print(lr1)
for lr2 in lanes_roles:
# create join key
tlr1 = '100' + '_' + lr1
tlr2 = '200' + '_' + lr2
data[tlr1 + '_' + tlr2] = data[tlr1].str.cat(data[tlr2], sep='_')
data = pd.merge(data,
h2h_win_rates[[
'champ_pair',
tlr1[4:] + '_' + tlr2[4:] + '_win_rate'
]],
how='left',
left_on=tlr1 + '_' + tlr2,
right_on='champ_pair')
data = data.rename(
{
tlr1[4:] + '_' + tlr2[4:] + '_win_rate':
tlr1 + '_' + tlr2 + '_h2h_100_wr'
},
axis=1)
data = data.drop(['champ_pair', tlr1 + '_' + tlr2], axis=1)
# Get rid of non-modeling columns
# cols_drop = ['match_id', 'game_version', 'queue_id', 'game_duration',
# '100_TOP_SOLO', '100_JUNGLE_NONE', '100_MIDDLE_SOLO', '100_BOTTOM_DUO_CARRY', '100_BOTTOM_DUO_SUPPORT',
# '200_TOP_SOLO', '200_JUNGLE_NONE', '200_MIDDLE_SOLO', '200_BOTTOM_DUO_CARRY', '200_BOTTOM_DUO_SUPPORT']
# data = data.drop(cols_drop)
train = data[data['match_id'].isin(train['match_id'])]
validation = data[data['match_id'].isin(validation['match_id'])]
test = data[data['match_id'].isin(test['match_id'])]
train.to_csv(train_file)
validation.to_csv(validation_file)
test.to_csv(test_file)
def main():
# Load match data
data = pd.read_csv(
os.getenv('MINED_DATA_DIR') + 'processed_match_data.csv')
# Train, validation, test split
train = data.sample(int(.6 * data.shape[0]), random_state=410)
validation = data[~data['match_id'].isin(train['match_id'])].sample(
int(.2 * data.shape[0]), random_state=411)
test = data[~data['match_id'].isin(
pd.concat([train['match_id'], validation['match_id']],
axis=0,
ignore_index=True))]
# Build win rates (optional)
# if len(sys.argv) > 1:
# indiv_wr_boolean = sys.argv[1]
# print('Fire 1')
# if len(sys.argv) > 2:
# paired_wr_boolean = sys.argv[2]
# print('Fire 2')
# if len(sys.argv) > 3:
# h2h_wr_boolean = sys.argv[3]
# print('Fire 3')
#build_win_rates(train, indiv_wr_boolean, paired_wr_boolean, h2h_wr_boolean)
# Load win rates
indiv_win_rates = pd.read_csv(indiv_win_rate_file)
paired_win_rates = pd.read_csv(paired_win_rate_file)
h2h_win_rates = pd.read_csv(h2h_win_rate_file)
indiv_win_rates = indiv_win_rates.rename({'Unnamed: 0': 'champ'},
axis='columns')
paired_win_rates = paired_win_rates.rename({'Unnamed: 0': 'champ_pair'},
axis='columns')
h2h_win_rates = h2h_win_rates.rename({'Unnamed: 0': 'champ_pair'},
axis='columns')
#### Win rate significance ####
# Filter out low games played win rates with win_rates.win_rate_significance()
#### Join win rates ####
teams_lanes_roles = dc.get_teams_lanes_roles()
# teams = dc.get_teams()
lanes_roles = dc.get_lanes_roles()
# Join individ win rates
for tlr in teams_lanes_roles:
# want to take index of indiv win rates and a single column of it
data = pd.merge(data,
indiv_win_rates[[
'champ', tlr[4:] + '_win_rate',
tlr[4:] + '_games_played'
]],
how='left',
left_on=tlr,
right_on='champ')
data = data.rename(
{
tlr[4:] + '_win_rate': tlr + '_wr',
tlr[4:] + '_games_played': tlr + '_gp'
},
axis=1)
data = data.drop(['champ'], axis=1)
# Join paired win rates
teams = dc.get_teams()
for lr1 in lanes_roles:
print(lr1)
for lr2 in lanes_roles:
if lr1 != lr2:
t2 = lr1 + '_' + lr2
paired_win_rates[t2 + '_wins'] = paired_win_rates[
t2 + '_rw'] + paired_win_rates[t2 + '_bw']
for team in teams:
# create join key
tlr1 = team + '_' + lr1
tlr2 = team + '_' + lr2
t4 = tlr1[4:] + '_' + tlr2[4:]
t41 = tlr1 + '_' + tlr2[4:]
data[tlr1 + '_' + tlr2] = data[tlr1].str.cat(data[tlr2],
sep='_')
data = pd.merge(data,
paired_win_rates[[
'champ_pair', t4 + '_wr', t4 + '_gp',
t4 + '_wins'
]],
how='left',
left_on=tlr1 + '_' + tlr2,
right_on='champ_pair')
data = data.rename(
{
t4 + '_wr': t41 + '_wr',
t4 + '_gp': t41 + '_gp',
t4 + '_wins': t41 + '_wins',
},
axis=1)
data = data.drop(['champ_pair', tlr1 + '_' + tlr2], axis=1)
# Join h2h win rates
for lr1 in lanes_roles:
print(lr1)
for lr2 in lanes_roles:
# create join key
tlr1 = '100' + '_' + lr1
tlr2 = '200' + '_' + lr2
data[tlr1 + '_' + tlr2] = data[tlr1].str.cat(data[tlr2], sep='_')
data = pd.merge(data,
h2h_win_rates[[
'champ_pair',
tlr1[4:] + '_' + tlr2[4:] + '_wr',
tlr1[4:] + '_' + tlr2[4:] + '_gp',
tlr1[4:] + '_' + tlr2[4:] + '_wins'
]],
how='left',
left_on=tlr1 + '_' + tlr2,
right_on='champ_pair')
data = data.rename(
{
tlr1[4:] + '_' + tlr2[4:] + '_wr':
tlr1 + '_' + tlr2 + '_wr',
tlr1[4:] + '_' + tlr2[4:] + '_gp':
tlr1 + '_' + tlr2 + '_gp',
tlr1[4:] + '_' + tlr2[4:] + '_wins':
tlr1 + '_' + tlr2 + '_wins'
},
axis=1)
data = data.drop(['champ_pair', tlr1 + '_' + tlr2], axis=1)
# Drop erroneous rows with '0' as a champ name
print(data.shape)
lane_cols = [
'100_TOP_SOLO', '100_JUNGLE_NONE', '100_MIDDLE_SOLO',
'100_BOTTOM_DUO_CARRY', '100_BOTTOM_DUO_SUPPORT', '200_TOP_SOLO',
'200_JUNGLE_NONE', '200_MIDDLE_SOLO', '200_BOTTOM_DUO_CARRY',
'200_BOTTOM_DUO_SUPPORT'
]
for col in lane_cols:
data = data[data[col] != '0']
print(data.shape)
train = data[data['match_id'].isin(train['match_id'])]
validation = data[data['match_id'].isin(validation['match_id'])]
test = data[data['match_id'].isin(test['match_id'])]
# Adjust the paired win rates to not include that record's game in the calculation
for lr1 in lanes_roles:
print(lr1)
for lr2 in lanes_roles:
if lr1 != lr2:
t2 = lr1 + '_' + lr2
for team in teams:
# create join key
tlr1 = team + '_' + lr1
tlr2 = team + '_' + lr2
t41 = tlr1 + '_' + tlr2[4:]
if team == '100':
train[t41 +
'_wins'] = train[t41 +
'_wins'] - train['team_100_win']
else:
train[t41 + '_wins'] = train[t41 + '_wins'] - (
1 - train['team_100_win'])
train[t41 + '_gp'] = train[t41 + '_gp'] - 1
train[t41 + '_gp'] = np.maximum(0, train[t41 + '_gp'])
train[t41 +
'_wr'] = train[t41 + '_wins'] / train[t41 + '_gp']
train[t41 + '_wr'] = np.maximum(0, train[t41 + '_wr'])
train[t41 + '_wr'].fillna(0, inplace=True)
# validation[t41 + '_wr'] = validation[t41 + '_wins'] / validation[t41 + '_gp']
# validation[t41 + '_wr'] = np.maximum(0, validation[t41 + '_wr'])
# validation[t41 + '_wr'].fillna(0, inplace=True)
# test[t41 + '_wr'] = test[t41 + '_wins'] / test[t41 + '_gp']
# test[t41 + '_wr'] = np.maximum(0, test[t41 + '_wr'])
# test[t41 + '_wr'].fillna(0, inplace=True)
# Adjust the paired h2h win rates to not include that record's game in the calculation
for lr1 in lanes_roles:
print(lr1)
for lr2 in lanes_roles:
tlr1 = '100' + '_' + lr1
tlr2 = '200' + '_' + lr2
t12 = tlr1 + '_' + tlr2
train[t12 + '_wins'] = train[t12 + '_wins'] - train['team_100_win']
train[t12 + '_gp'] = train[t12 + '_gp'] - 1
train[t12 + '_wr'] = train[t12 + '_wins'] / train[t12 + '_gp']
train[t12 + '_wr'] = np.maximum(0, train[t12 + '_wr'])
train[t12 + '_wr'].fillna(0, inplace=True)
cols_to_drop = [x for x in train.columns if x[-4:] == 'wins']
train.drop(cols_to_drop, inplace=True, axis=1)
validation.drop(cols_to_drop, inplace=True, axis=1)
test.drop(cols_to_drop, inplace=True, axis=1)
train.to_csv(train_file)
validation.to_csv(validation_file)
test.to_csv(test_file)