def player_team_elo_constant(self, data_total_info,total_players_id,k_factor=16):
dota_team_id=set(list(data_total_info.W_id)+list(data_total_info.L_id))
dota_team_id=pd.DataFrame(list(dota_team_id))
dota_team_id.index=dota_team_id[0]
dota_team_id['team_id']=range(len(dota_team_id))
W_team_id=[]
L_team_id=[]
mean_elo=1000.0
n_players = len(total_players_id)
n_teams = len(dota_team_id)
current_elos_player = np.ones(shape=(n_players))*mean_elo
current_elos_team = np.ones(shape=(n_teams))*mean_elo
current_count = np.zeros(n_teams)
for i in range(len(data_total_info)):
W_team_id.append(dota_team_id['team_id'][data_total_info['W_id'][i]])
L_team_id.append(dota_team_id['team_id'][data_total_info['L_id'][i]])
data_total_info['W_team_id']=W_team_id
data_total_info['L_team_id']=L_team_id
for row in data_total_info.itertuples():
idx = row.Index
# player elo
w_account1=row.W_players_1
l_account1=row.L_players_1
w_account2=row.W_players_2
l_account2=row.L_players_2
w_account3=row.W_players_3
l_account3=row.L_players_3
w_account4=row.W_players_4
l_account4=row.L_players_4
w_account5=row.W_players_5
l_account5=row.L_players_5
# tier = row.tier
# print(type(w_id))
# Get current elos
w_elo_before1 = current_elos_player[w_account1]
w_elo_before2 = current_elos_player[w_account2]
w_elo_before3 = current_elos_player[w_account3]
w_elo_before4 = current_elos_player[w_account4]
w_elo_before5 = current_elos_player[w_account5]
#print(w_elo_before1)
mean_w_elo_before = np.mean(
[int(w_elo_before1),
int(w_elo_before2),
int(w_elo_before3),
int(w_elo_before4),
int(w_elo_before5)
])
l_elo_before1 = current_elos_player[l_account1]
l_elo_before2 = current_elos_player[l_account2]
l_elo_before3 = current_elos_player[l_account3]
l_elo_before4 = current_elos_player[l_account4]
l_elo_before5 = current_elos_player[l_account5]
mean_l_elo_before = np.mean([
int(l_elo_before1),
int(l_elo_before2),
int(l_elo_before3),
int(l_elo_before4),
int(l_elo_before5)
])
# Update on game results
change_elo = elo.elo_score_constant(mean_w_elo_before,mean_l_elo_before,k_factor)
#change_elo = elo_player.elo_score_duration_diff(mean_w_elo_before, mean_l_elo_before,duration)
# Save updated elos
data_total_info.at[idx,'win_team_elo_player'] = mean_w_elo_before
data_total_info.at[idx,'Lose_team_elo_player'] = mean_l_elo_before
mean_w_elo_after, mean_l_elo_after = elo.update_elo_constant_k(mean_w_elo_before, mean_l_elo_before,k_factor)
w_elo_after1=w_elo_before1 + change_elo
l_elo_after1=l_elo_before1 - change_elo
w_elo_after2=w_elo_before2 + change_elo
l_elo_after2=l_elo_before2 - change_elo
w_elo_after3=w_elo_before3 + change_elo
l_elo_after3=l_elo_before3 - change_elo
w_elo_after4=w_elo_before4 + change_elo
l_elo_after4=l_elo_before4 - change_elo
w_elo_after5=w_elo_before5 + change_elo
l_elo_after5=l_elo_before5 - change_elo
current_elos_player[w_account1] = w_elo_after1
current_elos_player[l_account1] = l_elo_after1
current_elos_player[w_account2] = w_elo_after2
current_elos_player[l_account2] = l_elo_after2
current_elos_player[w_account3] = w_elo_after3
current_elos_player[l_account3] = l_elo_after3
current_elos_player[w_account4] = w_elo_after4
current_elos_player[l_account4] = l_elo_after4
current_elos_player[w_account5] = w_elo_after5
current_elos_player[l_account5] = l_elo_after5
# 如果新的战队之前没有出现过，通过选手的elo评分来给队伍定下基准分
w_id = row.W_team_id
l_id = row.L_team_id
w_elo_before = current_elos_team[w_id]
l_elo_before = current_elos_team[l_id]
w_team_times = current_count[w_id]
l_team_times = current_count[l_id]
w_team_times, l_team_times = w_team_times+1 ,l_team_times+1 # 计算出战队到最新打了多少次比赛
current_count[w_id] = w_team_times
current_count[l_id] = l_team_times
w_elo_after, l_elo_after = elo.update_elo_constant_k(w_elo_before, l_elo_before, k_factor)
if current_elos_team[w_id] == mean_elo and current_count[w_id] == 1.0:
current_elos_team[w_id] = mean_w_elo_after
data_total_info.at[idx,'win_team_elo_team'] = mean_w_elo_before
else:
current_elos_team[w_id] = w_elo_after
data_total_info.at[idx,'win_team_elo_team'] = w_elo_before
if current_elos_team[l_id] == mean_elo and current_count[l_id] == 1.0:
current_elos_team[l_id] = mean_l_elo_after
data_total_info.at[idx,'Lose_team_elo_team'] = mean_l_elo_before
else:
current_elos_team[l_id] = l_elo_after
data_total_info.at[idx,'Lose_team_elo_team'] = l_elo_before
data_total_info['dire_team_id']= data_total_info.dire_team
data_total_info['radiant_team_id'] = data_total_info.radiant_team
## win_percent
expected_list = []
for row in data_total_info.itertuples():
w_elo = row.win_team_elo_team
l_elo = row.Lose_team_elo_team
w_expected = elo.expected_result(w_elo, l_elo)
expected_list.append(w_expected)
data_total_info['win_predict'] = expected_list
data_total_info['lose_predict'] = 1-data_total_info.win_predict
result=[]
for row in data_total_info.itertuples():
if row.win_predict>0.5:
result.append(1)
else:
result.append(0)
data_total_info['result']=result
return(data_total_info)
def player_team_elo_result(self, data_total_info, total_players_id, k_factor=16):
dota_team_id = set(list(data_total_info.W_id) + list(data_total_info.L_id))
dota_team_id = pd.DataFrame(list(dota_team_id))
dota_team_id.index = dota_team_id[0]
dota_team_id['team_id'] = range(len(dota_team_id))
data_total_info['win_team_last_playes'] = data_total_info['W_account']
data_total_info['lose_team_last_playes'] = data_total_info['L_account']
W_team_id = []
L_team_id = []
mean_elo = 1000.0
n_players = len(total_players_id)
n_teams = len(dota_team_id)
current_elos_player = np.ones(shape=(n_players)) * mean_elo
current_elos_team = np.ones(shape=(n_teams)) * mean_elo
current_elos_combine = np.ones(shape=(n_teams)) * mean_elo # 结合队伍ELO与队员ELO分数
current_count = np.zeros(n_teams) #队伍比赛场次统计
current_count_player = np.zeros(n_players) #队员比赛场次统计
one_month_team_played = np.zeros(n_teams)
team_playes_id = [[0]] * n_teams# 设置储存上一次比赛的队伍的队员，通过比较与下一次的队伍队员是否有变化来判断出最新的队伍更换队员的 时间
current_team_playes_time = np.zeros(n_teams)
for i in range(len(data_total_info)):
W_team_id.append(dota_team_id['team_id'][data_total_info['W_id'][i]])
L_team_id.append(dota_team_id['team_id'][data_total_info['L_id'][i]])
data_total_info['W_team_id'] = W_team_id
data_total_info['L_team_id'] = L_team_id
for row in data_total_info.itertuples():
idx = row.Index
current = int(time.time())
# player elo
w_account1 = row.W_players_1
l_account1 = row.L_players_1
w_account2 = row.W_players_2
l_account2 = row.L_players_2
w_account3 = row.W_players_3
l_account3 = row.L_players_3
w_account4 = row.W_players_4
l_account4 = row.L_players_4
w_account5 = row.W_players_5
l_account5 = row.L_players_5
# tier = row.tier
# print(type(w_id))
# Get current elos
w_elo_before1 = current_elos_player[w_account1]
w_elo_before2 = current_elos_player[w_account2]
w_elo_before3 = current_elos_player[w_account3]
w_elo_before4 = current_elos_player[w_account4]
w_elo_before5 = current_elos_player[w_account5]
# print(w_elo_before1)
mean_w_elo_before = np.mean(
[int(w_elo_before1),
int(w_elo_before2),
int(w_elo_before3),
int(w_elo_before4),
int(w_elo_before5)
])
l_elo_before1 = current_elos_player[l_account1]
l_elo_before2 = current_elos_player[l_account2]
l_elo_before3 = current_elos_player[l_account3]
l_elo_before4 = current_elos_player[l_account4]
l_elo_before5 = current_elos_player[l_account5]
mean_l_elo_before = np.mean([
int(l_elo_before1),
int(l_elo_before2),
int(l_elo_before3),
int(l_elo_before4),
int(l_elo_before5)
])
# Update on game results
change_elo = elo.elo_score_constant(mean_w_elo_before, mean_l_elo_before, k_factor)
# change_elo = elo_player.elo_score_duration_diff(mean_w_elo_before, mean_l_elo_before,duration)
# Save updated elos
data_total_info.at[idx, 'win_team_elo_player'] = mean_w_elo_before
data_total_info.at[idx, 'Lose_team_elo_player'] = mean_l_elo_before
mean_w_elo_after, mean_l_elo_after = elo.update_elo_constant_k(mean_w_elo_before, mean_l_elo_before,
k_factor)
w_elo_after1 = w_elo_before1 + change_elo
l_elo_after1 = l_elo_before1 - change_elo
w_elo_after2 = w_elo_before2 + change_elo
l_elo_after2 = l_elo_before2 - change_elo
w_elo_after3 = w_elo_before3 + change_elo
l_elo_after3 = l_elo_before3 - change_elo
w_elo_after4 = w_elo_before4 + change_elo
l_elo_after4 = l_elo_before4 - change_elo
w_elo_after5 = w_elo_before5 + change_elo
l_elo_after5 = l_elo_before5 - change_elo
current_elos_player[w_account1] = w_elo_after1
current_elos_player[l_account1] = l_elo_after1
current_elos_player[w_account2] = w_elo_after2
current_elos_player[l_account2] = l_elo_after2
current_elos_player[w_account3] = w_elo_after3
current_elos_player[l_account3] = l_elo_after3
current_elos_player[w_account4] = w_elo_after4
current_elos_player[l_account4] = l_elo_after4
current_elos_player[w_account5] = w_elo_after5
current_elos_player[l_account5] = l_elo_after5
# 如果新的战队之前没有出现过，通过选手的elo评分来给队伍定下基准分
w_id = row.W_team_id
l_id = row.L_team_id
W_account = row.W_account
L_account = row.L_account
start_time = row.start_time
w_elo_before = current_elos_team[w_id]
l_elo_before = current_elos_team[l_id]
w_team_times = current_count[w_id]
l_team_times = current_count[l_id]
'''
计算出每个选手参加的比赛场次
'''
w_player_1_times = current_count_player[w_account1]
w_player_2_times = current_count_player[w_account2]
w_player_3_times = current_count_player[w_account3]
w_player_4_times = current_count_player[w_account4]
w_player_5_times = current_count_player[w_account5]
l_player_1_times = current_count_player[l_account1]
l_player_2_times = current_count_player[l_account2]
l_player_3_times = current_count_player[l_account3]
l_player_4_times = current_count_player[l_account4]
l_player_5_times = current_count_player[l_account5]
w_team_times, l_team_times = w_team_times + 1, l_team_times + 1 # 计算出战队到最新打了多少次比赛
w_player_1_times, w_player_2_times, w_player_3_times, w_player_4_times, w_player_5_times = w_player_1_times+1,w_player_2_times+1, w_player_3_times+1, w_player_4_times+1, w_player_5_times+1
l_player_1_times, l_player_2_times, l_player_3_times, l_player_4_times, l_player_5_times = l_player_1_times+1,l_player_2_times+1, l_player_3_times+1, l_player_4_times+1, l_player_5_times+1
current_count_player[w_account1] = w_player_1_times
current_count_player[w_account2] = w_player_2_times
current_count_player[w_account3] = w_player_3_times
current_count_player[w_account4] = w_player_4_times
current_count_player[w_account5] = w_player_5_times
current_count_player[l_account1] = l_player_1_times
current_count_player[l_account2] = l_player_2_times
current_count_player[l_account3] = l_player_3_times
current_count_player[l_account4] = l_player_4_times
current_count_player[l_account5] = l_player_5_times
current_count[w_id] = w_team_times
current_count[l_id] = l_team_times
if len(set(team_playes_id[w_id]) - set(W_account)) != 0:
team_playes_id[w_id] = W_account
current_team_playes_time[w_id] = start_time
else:
pass
if len(set(team_playes_id[l_id]) - set(L_account)) != 0:
team_playes_id[l_id] = L_account
current_team_playes_time[l_id] = start_time
else:
pass
w_elo_after, l_elo_after = elo.update_elo_constant_k(w_elo_before, l_elo_before, k_factor)
if start_time > (current - 2592000):
one_month_team_played[w_id] = one_month_team_played[w_id] + 1
one_month_team_played[l_id] = one_month_team_played[l_id] + 1
data_total_info.at[idx, 'one_month_team_playerd_w_team'] = one_month_team_played[w_id]
data_total_info.at[idx, 'one_month_team_playerd_l_team'] = one_month_team_played[l_id]
else:
pass
if current_elos_team[w_id] == mean_elo and current_count[w_id] == 1.0:
current_elos_team[w_id] = mean_w_elo_after
data_total_info.at[idx, 'win_team_elo_team'] = mean_w_elo_before
else:
current_elos_team[w_id] = w_elo_after
data_total_info.at[idx, 'win_team_elo_team'] = w_elo_before
if current_elos_team[l_id] == mean_elo and current_count[l_id] == 1.0:
current_elos_team[l_id] = mean_l_elo_after
data_total_info.at[idx, 'Lose_team_elo_team'] = mean_l_elo_before
else:
current_elos_team[l_id] = l_elo_after
data_total_info.at[idx, 'Lose_team_elo_team'] = l_elo_before
## 如果选手评分ELO与队伍评分ELO差距较大的时候 采用选手评分ELO
if abs(mean_w_elo_after - w_elo_after)>=100:
current_elos_combine[w_id] = mean_w_elo_after
elif abs(mean_w_elo_after - w_elo_after) >50 and abs(mean_w_elo_after - w_elo_after) < 100:
current_elos_combine[w_id] = 0.65* mean_w_elo_after + 0.35*w_elo_after
elif abs(mean_w_elo_after- w_elo_after) >25 and abs(mean_w_elo_after- w_elo_after) < 50:
current_elos_combine[w_id] = 0.35 * mean_w_elo_after + 0.65*w_elo_after
else:
current_elos_combine[w_id] = w_elo_after
if abs(mean_l_elo_after - l_elo_after)>=100:
current_elos_combine[l_id] = mean_w_elo_after
elif abs(mean_l_elo_after - l_elo_after) >50 and abs(mean_l_elo_after - l_elo_after) < 100:
current_elos_combine[l_id] = 0.65* mean_l_elo_after + 0.35*l_elo_after
elif abs(mean_l_elo_after- l_elo_after) >25 and abs(mean_l_elo_after- l_elo_after) < 50:
current_elos_combine[l_id] = 0.35 * mean_l_elo_after + 0.65*l_elo_after
else:
current_elos_combine[l_id] = l_elo_after
data_total_info.at[idx, 'win_team_ELO_combine'] = current_elos_combine[w_id]
data_total_info.at[idx, 'lose_team_ELO_combine'] = current_elos_combine[l_id]
data_total_info.at[idx, 'win_team_last_playes'] = team_playes_id[w_id]
data_total_info.at[idx, 'lose_team_last_playes'] = team_playes_id[l_id]
data_total_info['dire_team_id'] = data_total_info.dire_team
data_total_info['radiant_team_id'] = data_total_info.radiant_team
## win_percent
expected_list_team = []
expected_list_player =[]
for row in data_total_info.itertuples():
w_elo_team = row.win_team_elo_team
l_elo_team = row.Lose_team_elo_team
w_elo_player = row.win_team_elo_player
l_elo_player = row.Lose_team_elo_player
w_expected_team = elo.expected_result(w_elo_team, l_elo_team)
w_expected_plyaer =elo.expected_result(w_elo_player, l_elo_player)
expected_list_team.append(w_expected_team)
expected_list_player.append(w_expected_plyaer)
data_total_info['win_predict_team'] = expected_list_team
data_total_info['win_predict_player'] = expected_list_player
data_total_info['lose_predict_team'] = 1 - data_total_info.win_predict_team
data_total_info['lose_predict_player'] = 1 - data_total_info.win_predict_player
result_player = []
result_team = []
for row in data_total_info.itertuples():
if row.win_predict_team > 0.5:
result_team.append(1)
else:
result_team.append(0)
if row.win_predict_player > 0.5:
result_player.append(1)
else:
result_player.append(0)
data_total_info['result_team'] = result_team
data_total_info['result_player'] = result_player
return (current_elos_combine, current_count, current_team_playes_time, one_month_team_played,current_elos_player,current_count_player)