def __init__(self, config):
self.salary_cap = 60000
self.selection_multiplier = 100 #Multiply EFPPG by 100 so we can encapsulate remainder in integer array
self.config = config
self.team_selector = TeamSelector(config['TeamSelector'])
class TeamGenerator:
def __init__(self, config):
self.salary_cap = 60000
self.selection_multiplier = 100 #Multiply EFPPG by 100 so we can encapsulate remainder in integer array
self.config = config
self.team_selector = TeamSelector(config['TeamSelector'])
def select_top_team(self, player_stats, player_costs, date):
print 'Selecting top team for date: ' + str(date)
start_time = time.time()
players = self._merge_player_stats_and_costs(player_stats, player_costs, date)
players = self._filter_players(players)
possible_teams = self._all_possible_teams(players)
possible_teams = self._filter_on_cap_exceeded_np(possible_teams)
top_team = self.team_selector.select_team(possible_teams, player_stats, date)
end_time = time.time()
print 'Simulation finished with runtime: ' + str(end_time - start_time)
expected_points = self._expected_points(top_team)
self._print_top_teams(top_team, players)
return {'team': top_team, 'expected': expected_points}
def _merge_player_stats_and_costs(self, player_stats, player_costs, date):
player_stats['Name'] = player_stats['PLAYER']
player_stats = player_stats[player_stats['Date'] < date]
player_stats = player_stats.sort_values('Date')
player_stats = player_stats.groupby('Player_ID').tail(1)
player_costs['Name'] = player_costs['First Name'] + ' ' + player_costs['Last Name']
player_costs['FPPG_Per_Dollar'] = player_costs['FPPG'] / player_costs['Salary'] * 10000
player_stats_costs = pd.merge(player_costs, player_stats, on='Name', how='outer')
# Eliminate players with no E_FPPG
player_stats_costs = player_stats_costs[np.isfinite(player_stats_costs[self.config['Selection_Field']])]
player_stats_costs['EFPPG_Per_Dollar'] = player_stats_costs['E_FPPG'] / player_stats_costs['Salary'] *10000
player_stats_costs = player_stats_costs[np.isfinite(player_stats_costs['EFPPG_Per_Dollar'])]
print 'Unfiltered Player Count : {0}'.format(len(player_costs))
return player_stats_costs
def _filter_players(self, players):
filtered_players = players[players['Injury Indicator'].isnull()]
filtered_players = filtered_players.sort_values(self.config['Filter_Field'], ascending=False)
filtered_players = filtered_players.iloc[0:self.config['Player_Count']]
print 'Filtered Player Count : {0}'.format(len(filtered_players))
return filtered_players
def _filter_team_fppg_per_dollar_pctile(self, input_teams, pctile):
fppg_totals = np.sum(input_teams[:,:,2], axis=1) / (1.0 * np.sum(input_teams[:,:,3], axis=1))
new_len = int(input_teams.shape[0]*pctile)
index = np.argsort(fppg_totals)
return input_teams[index[new_len:], :, :]
def _filter_on_cap_exceeded_np(self, input_teams):
mask = np.sum(input_teams[:, :, 3], axis=1) <= self.salary_cap
return input_teams[mask, :, :]
def _cartesian(self, x, y):
return np.concatenate((np.tile(x, (len(y), 1, 1)), np.repeat(y, len(x), axis=0)), axis=1)
def _player_array_from_position(self, players, position):
mask = players['Position'] == position
player_data = [players['Player_ID'][mask], players['TeamID'][mask], players[self.config['Selection_Field']][mask] * self.selection_multiplier, players['Salary'][mask]]
player_data = np.asarray(player_data).astype(np.int32).T.tolist()
return player_data
def _possible_teams_from_combos(self, first_position_combos, second_position_combos, filter_size_bytes):
first_position_combos = self._filter_team_fppg_per_dollar_pctile(first_position_combos, filter_size_bytes)
return self._cartesian(first_position_combos, second_position_combos)
def _position_combos(self, players, position, num):
position_players = self._player_array_from_position(players, position)
if num == 1:
return np.expand_dims(np.array(position_players), 1).astype(np.int32)
else:
return np.array(list(combinations(position_players, 2))).astype(np.int32)
def _all_possible_teams(self, players):
pg_combos = self._position_combos(players, 'PG', 2)
sg_combos = self._position_combos(players, 'SG', 2)
pf_combos = self._position_combos(players, 'PF', 2)
sf_combos = self._position_combos(players, 'SF', 2)
c_combos = self._position_combos(players, 'C', 1)
possible_teams = self._possible_teams_from_combos(pg_combos, sg_combos, self.config['Team_Size_Bytes_1'])
possible_teams = self._possible_teams_from_combos(possible_teams, pf_combos, self.config['Team_Size_Bytes_1'])
possible_teams = self._possible_teams_from_combos(possible_teams, sf_combos, self.config['Team_Size_Bytes_2'])
return self._possible_teams_from_combos(possible_teams, c_combos, self.config['Team_Size_Bytes_3'])
def _expected_points(self, team):
return np.sum(team[:,:,2], axis=1) / (1.0 * self.selection_multiplier)
def _print_top_teams(self, top_teams, players):
fppg_totals = np.sum(top_teams[:,:,2], axis=1) / (1.0 * self.selection_multiplier)
salary_totals = np.sum(top_teams[:, :, 3], axis=1)
for i, team in enumerate(top_teams):
print "Team Expectation: " + str(fppg_totals[i])
print "Salary Utilized: " + str(salary_totals[i])
for player_id in team[:,0]:
if player_id > 0:
player = players['Name'][players['Player_ID'] == player_id].iloc[0]
team = players['Team'][players['Player_ID'] == player_id].iloc[0]
position = players['Position'][players['Player_ID'] == player_id].iloc[0]
points = players[self.config['Selection_Field']][players['Player_ID']==player_id].iloc[0]
salary = players['Salary'][players['Player_ID']==player_id].iloc[0]
print player + ' ' + team + ' ' + position + ' ' + str(points) + ' $' + str(salary)
print ''
