def generate_elo(df,counts_i):
players_list = np.union1d(df.w_name, df.l_name)
players_elo = dict(zip(list(players_list), [elo.Rating() for __ in range(len(players_list))]))
surface_elo = {}
for surface in ('Hard','Clay','Grass'):
surface_elo[surface] = dict(zip(list(players_list), [elo.Rating() for __ in range(len(players_list))]))
elo_1s, elo_2s = [],[]
surface_elo_1s, surface_elo_2s = [],[]
elo_obj = elo.Elo_Rater()
# update player elo from every recorded match
for i, row in df.iterrows():
surface = row['surface']; is_gs = row['is_gs']
# append elos, rate, update
w_elo,l_elo = players_elo[row['w_name']],players_elo[row['l_name']]
elo_1s.append(w_elo.value);elo_2s.append(l_elo.value)
elo_obj.rate_1vs1(w_elo,l_elo,is_gs,counts=counts_i)
surface_elo_1s.append(surface_elo[surface][row['w_name']].value if surface in ('Hard','Clay','Grass') else w_elo.value)
surface_elo_2s.append(surface_elo[surface][row['l_name']].value if surface in ('Hard','Clay','Grass') else l_elo.value)
if surface in ('Hard','Clay','Grass'):
new_elo1, new_elo2 = elo_obj.rate_1vs1(surface_elo[surface][row['w_name']],surface_elo[surface][row['l_name']],is_gs,counts=counts_i)
# add columns
if counts_i:
df['w_elo_538'], df['l_elo_538'] = elo_1s, elo_2s
df['w_sf_elo_538'], df['l_sf_elo_538'] = surface_elo_1s, surface_elo_2s
else:
df['w_elo'], df['l_elo'] = elo_1s, elo_2s
df['w_sf_elo'], df['l_sf_elo'] = surface_elo_1s, surface_elo_2s
return df
def generate_elo_columns(arr, counts_538):
# print('arr[0]: ', arr[0])
# print('arr[1]: ', arr[1])
# print('arr[0].dtype', arr[:, 0].dtype)
# print('arr[1].dtype', arr[:, 1].dtype)
# print('arr[:, :2]', arr[:, :2])
# for s in arr:
# if isinstance(s[1], bool):
# print('s: ', s)
# print('is: ', [s for s in arr if isinstance(s[0], bool))
# players_set = np.unique(arr[:, :2].astype(str))
player_names = arr[:, :2].flatten()
players_set = np.where(player_names!=player_names, '', player_names).tolist()
# players_set = list(set(list(np.concatenate(arr[:, 0], arr[:, 1]))))
# player_count = len(players_set)
# print('player_count: ', player_count)
# initial_ratings = [elo.Rating() for _ in range(player_count)]
# zipped = zip(
# players_set,
# [elo.Rating() for _ in range(player_count)]
# )
# # print('zipped: ', zipped)
# players_elo = dict(zip(
# players_set,
# [elo.Rating() for _ in range(player_count)]
# )) # can use default dict here?
players_elo = {}
for player in players_set:
# print('player: ', player)
players_elo[player] = elo.Rating()
match_elos = np.zeros([arr.shape[0], 2])
elo_obj = elo.Elo_Rater()
# update player elo from every recorded match
for i in range(arr.shape[0]):
w_name, l_name = arr[i][:2]
if w_name != w_name or l_name != l_name:
match_elos[i] = np.nan, np.nan
continue
match_elos[i] = players_elo[w_name].value, players_elo[l_name].value
elo_obj.rate_1vs1(players_elo[w_name], players_elo[l_name], arr[i][2], counts_538)
return match_elos[:,0], match_elos[:,1]
def generateEloColumnsWithHistory(arr, counts_538):
playerEloHistory = defaultdict(list)
players_set = np.unique(arr[:, :2])
players_elo = dict(zip(
players_set,
[elo.Rating() for __ in range(len(players_set))]
)) # can use default dict here?
match_elos = np.zeros([arr.shape[0], 2])
elo_obj = elo.Elo_Rater()
# update player elo from every recorded match
for i in range(arr.shape[0]):
w_name, l_name = arr[i][:2]
isGrandSlam = arr[i][2]
date = datetime.datetime.strptime(arr[i][3], '%Y-%m-%d')
match_elos[i] = players_elo[w_name].value, players_elo[l_name].value
elo_obj.rate_1vs1(players_elo[w_name], players_elo[l_name], 0, counts_538)
playerEloHistory[w_name].append({ 'date': date, 'newElo': players_elo[w_name].value, 'won': 1 })
playerEloHistory[l_name].append({ 'date': date, 'newElo': players_elo[l_name].value, 'won': 0 })
return match_elos[:,0], match_elos[:,1], playerEloHistory, players_elo
def generate_elo_stephanie(df, counts, h1, h2):
players_list = np.union1d(df.w_name, df.l_name)
players_elo = dict(
zip(list(players_list),
[elo.Rating() for __ in range(len(players_list))]))
surface_elo = {}
for surface in ('Hard', 'Clay', 'Grass'):
surface_elo[surface] = dict(
zip(list(players_list),
[elo.Rating() for __ in range(len(players_list))]))
elo_1s, elo_2s = [], []
surface_elo_1s, surface_elo_2s = [], []
elo_obj = elo.Elo_Rater()
k1, k2 = 5.3, 16
# update player elo from every recorded match
for i, row in df.iterrows():
surface = row['surface']
is_gs = row['is_gs']
tny_name = row['tny_name']
tny_round_name = row['tourney_round_name']
#delta1 = row['newwdelta1'] ; delta2 = row['newwdelta2']
delta1 = row['minusnewwdelta1']
delta2 = row['minusnewwdelta2']
# append elos, rate, update
w_elo, l_elo = players_elo[row['w_name']], players_elo[row['l_name']]
elo_1s.append(w_elo.value)
elo_2s.append(l_elo.value)
# get different kinds of precalculated scalers
# s_tournament_w = row['s_tournament_w']
# s_tournament_l = row['s_tournament_l']
# s_match_duration_w = row['s_match_duration_w']
# s_match_duration_l = row['s_match_duration_l']
# tny_name = row['tny_name']
# tny_round_name = row['tourney_round_name']
# is_gs = row['is_gs']
# surface = row['surface']
# w_name = row['w_name']
# l_name = row['l_name']
elo_obj.rate_1vs1_stephanie(w_elo,
l_elo,
k1,
k2,
delta1,
delta2,
is_gs,
counts,
tny_name,
tny_round_name,
h1=h1,
h2=h2)
surface_elo_1s.append(surface_elo[surface][row['w_name']].
value if surface in ('Hard', 'Clay',
'Grass') else w_elo.value)
surface_elo_2s.append(surface_elo[surface][row['l_name']].
value if surface in ('Hard', 'Clay',
'Grass') else l_elo.value)
if surface in ('Hard', 'Clay', 'Grass'):
elo_obj.rate_1vs1_stephanie(w_elo,
l_elo,
k1,
k2,
delta1,
delta2,
is_gs,
counts,
tny_name,
tny_round_name,
h1=h1,
h2=h2)
# add columns
if counts:
df['w_elo_538'], df['l_elo_538'] = elo_1s, elo_2s
df['w_sf_elo_538'], df['l_sf_elo_538'] = surface_elo_1s, surface_elo_2s
else:
df['w_elo'], df['l_elo'] = elo_1s, elo_2s
df['w_sf_elo'], df['l_sf_elo'] = surface_elo_1s, surface_elo_2s
return df
def generate_elo(df, counts_i):
players_list = np.union1d(df.w_name, df.l_name)
players_elo = dict(
zip(list(players_list),
[elo.Rating() for __ in range(len(players_list))]))
surface_elo = {}
for surface in ('Hard', 'Clay', 'Grass'):
surface_elo[surface] = dict(
zip(list(players_list),
[elo.Rating() for __ in range(len(players_list))]))
elo_1s, elo_2s = [], []
surface_elo_1s, surface_elo_2s = [], []
elo_obj = elo.Elo_Rater()
# update player elo from every recorded match
for i, row in df.iterrows():
# if surface in ('Hard','Clay','Grass'):
# print "same object: ", players_elo[row['w_name']] is surface_elo[surface][row['w_name']]
surface = row['surface']
is_gs = row['is_gs']
# append elos, rate, update
w_elo, l_elo = players_elo[row['w_name']], players_elo[row['l_name']]
elo_1s.append(w_elo.value)
elo_2s.append(l_elo.value)
tny_name = row['tny_name']
tny_round_name = row['tourney_round_name']
# get different kinds of precalculated scalers
# s_tournament_w = row['s_tournament_w']
# s_tournament_l = row['s_tournament_l']
# s_match_duration_w = row['s_match_duration_w']
# s_match_duration_l = row['s_match_duration_l']
# w_name = row['w_name']
# l_name = row['l_name']
# original
elo_obj.rate_1vs1(w_elo, l_elo, is_gs, counts=counts_i)
# elo_obj.rate_1vs1(w_elo,l_elo,is_gs, counts=counts_i, tny_name=tny_name, tny_round_name=tny_round_name)
surface_elo_1s.append(surface_elo[surface][row['w_name']].
value if surface in ('Hard', 'Clay',
'Grass') else w_elo.value)
surface_elo_2s.append(surface_elo[surface][row['l_name']].
value if surface in ('Hard', 'Clay',
'Grass') else l_elo.value)
if surface in ('Hard', 'Clay', 'Grass'):
elo_obj.rate_1vs1(surface_elo[surface][row['w_name']],
surface_elo[surface][row['l_name']],
is_gs,
counts=counts_i)
# elo_obj.rate_1vs1(surface_elo[surface][row['w_name']],surface_elo[surface][row['l_name']], counts=counts_i, tny_name=tny_name, tny_round_name=tny_round_name)
# add columns
if counts_i:
df['w_elo_538'], df['l_elo_538'] = elo_1s, elo_2s
df['w_sf_elo_538'], df['l_sf_elo_538'] = surface_elo_1s, surface_elo_2s
else:
df['w_elo'], df['l_elo'] = elo_1s, elo_2s
df['w_sf_elo'], df['l_sf_elo'] = surface_elo_1s, surface_elo_2s
return df