def setUp(self):
"""Initialization"""
self.db_root = Config().data_dir
self.repoCode = get_git_root()
self.repoModel = path.join(
self.repoCode,
'ReinforcementLearning/NHL/playerstats/offVSdef/Automatic_classification/MODEL_perceptron_1layer_10units_relu'
)
def setUp(self):
"""Initialization"""
self.db_root = Config(
).data_dir # This is the location of the Hockey database
self.repoCode = get_git_root()
self.repoModel = path.join(
self.repoCode,
'ReinforcementLearning/NHL/playerstats/offVSdef/Automatic_classification/MODEL_perceptron_1layer_10units_relu'
)
# Now lets get game data
self.season = Season(
db_root=self.db_root, repo_model=self.repoModel,
year_begin=2012) # Season.from_year_begin(2012) # '20122013'
# Montreal received Ottawa on march 13, 2013, let's convert game date to game code
gameId = self.season.get_game_id(home_team_abbr='MTL',
game_date=datetime.date(year=2013,
month=3,
day=13))
self.a_game = Game(self.season, gameId=gameId)
graph = classifier['sess'].graph
classifier['annX'] = graph.get_tensor_by_name('Input_to_the_network-player_features:0')
classifier['annY'] = graph.get_tensor_by_name('prediction:0')
saver.restore(classifier['sess'], tf.train.latest_checkpoint(path.join(repoModel, './')))
return players_model, classifier
# ========
# LAUNCHER
# ========
# === Set pointers
# Pointers to the data
repoCode = get_git_root()
db_root = Config().data_dir #This is the location of the Hockey database
# db_root = '/Users/younes_zerouali/Documents/Stradigi/Databases/Hockey'
# db_root = '/Users/luisd/dev/NHL_stats/data'
repoPbP = path.join(db_root, 'PlayByPlay')
repoPSt = path.join(db_root, 'PlayerStats/player')
repoModel = path.join(repoCode, 'ReinforcementLearning/NHL/playerstats/offVSdef/Automatic_classification/MODEL_backup_trainedonallseasons_rankstatprediction')
repoModel = path.join(repoCode, 'ReinforcementLearning/NHL/playerstats/offVSdef/Automatic_classification/MODEL_perceptron_1layer_10units_relu')
repoSave = None #path.join(repoCode, 'ReinforcementLearning/NHL/playbyplay/data')
# === Pipeline
# Select a season
season = '20122013'
# List all games
HSS = HockeySS(repoPbP, repoPSt)
HSS.list_all_games()
if __name__ == '__main__':
from Utils.base import get_logger
import pandas as pd
import matplotlib.pyplot as plt
import os
from ReinforcementLearning.NHL.config import Config
common_logger = get_logger(name="common_logger",
debug_log_file_name="common_logger.log")
print("Debug will be written in {}".format(
common_logger.handlers[1].baseFilename))
coaches_evaluator = Coach_Evaluator(common_logger)
grades = coaches_evaluator.read_and_display_coaches_evals()
standings_file_name = os.path.join(Config().data_dir,
"NHLStandings2012-2013.csv")
df = pd.DataFrame.from_csv(standings_file_name)
# add column with abbreviation
df['ABBREV'] = pd.Series(list(map(name2abbrev, df["TEAM"].values)),
index=df.index)
df['GRADE'] = pd.Series(list(
map(lambda abbrev: grades[abbrev], df["ABBREV"].values)),
index=df.index)
df['WINS'] = pd.Series(list(
map(lambda record: int(record.split("-")[0]), df["Overall"].values)),
index=df.index)
df.plot(kind='scatter', x="GRADE", y="WINS") # , 'rx')
plt.show()
def evaluate_all_coaches(self, season_year_begin: int,
teams_opt: Optional[List[str]], n_games: int):
from ReinforcementLearning.NHL.playbyplay.state_space_data import HockeySS
"""Initialization"""
os.makedirs(self.base_dir, exist_ok=True)
my_config = Config()
self.alogger.debug("Data configured to be in '%s'" %
(my_config.data_dir))
db_root = my_config.data_dir
repoCode = get_git_root()
repoModel = path.join(
repoCode,
'ReinforcementLearning/NHL/playerstats/offVSdef/Automatic_classification/MODEL_perceptron_1layer_10units_relu'
)
season = Season(self.alogger,
db_root=db_root,
year_begin=season_year_begin,
repo_model=repoModel)
# Line translation table
linedict = HockeySS(db_root)
linedict.make_line_dictionary()
linedict = linedict.line_dictionary
# Load the Qvalues table
Qvalues = \
pickle.load(open(path.join(repoCode, 'ReinforcementLearning/NHL/playbyplay/data/stable/RL_action_values.p'), 'rb'))[
'action_values']
# Visualize it dimensions (period x differential x away line's code x home line's code)
print('Q-table dimensions: ', Qvalues.shape)
# for what teams will we run this calculation?
calc_teams = season.get_teams() if teams_opt is None else teams_opt
for a_team in calc_teams:
season.alogger.debug("=============> calculating %s" % (a_team))
seconds = get_teams_coach_performance(season,
team_abbr=a_team,
maybe_a_starting_date=None,
line_dict=linedict,
Qvalues=Qvalues,
how_many_games=n_games)
season.alogger.debug(seconds)
if seconds["does_not_match_optimal"] == seconds[
"matches_optimal"] == 0:
season.alogger.info(
"[team: '%s'] No evidence for coach to be evaluated on." %
(a_team))
else:
total_secs = seconds['matches_optimal'] + seconds[
'does_not_match_optimal']
season.alogger.info(
"['%s'] Home coach's score is %d (secs. optimal) / %d (secs. total) = %.2f (in [0,1])"
% (a_team, seconds['matches_optimal'], total_secs,
seconds['matches_optimal'] / total_secs))
file_to_save = os.path.join(self.base_dir, a_team + ".pkl")
self.alogger.debug("Saving data for '%s' in file %s" %
(a_team, file_to_save))
with open(file_to_save, 'wb') as dict_file:
pickle.dump(seconds, dict_file)
self.alogger.debug("DONE")
def __init__(self, alogger: logging.Logger):
self.alogger = alogger
self.base_dir = os.path.join(Config().data_dir, "coaches_perf")
self.alogger.debug("Read info from '%s'" % (self.base_dir))
from ReinforcementLearning.NHL.lines.coach import Coach_Evaluator
from ReinforcementLearning.NHL.lines.correlation_with_performance import name2abbrev
if __name__ == '__main__':
from Utils.base import get_logger
import pandas as pd
import matplotlib.pyplot as plt
import os
from ReinforcementLearning.NHL.config import Config
common_logger = get_logger(name="common_logger", debug_log_file_name="common_logger.log")
print("Debug will be written in {}".format(common_logger.handlers[1].baseFilename))
coaches_evaluator = Coach_Evaluator(common_logger)
grades = coaches_evaluator.read_and_display_coaches_evals()
attendance_file_name = os.path.join(Config().data_dir, "NHLAttendance2012-2013.csv")
df = pd.DataFrame.from_csv(attendance_file_name)
# add column with abbreviation
df['ABBREV'] = pd.Series(list(map(name2abbrev, df["TEAM"].values)), index=df.index)
df['GRADE'] = pd.Series(list(map(lambda abbrev: grades[abbrev], df["ABBREV"].values)), index=df.index)
df.AVG = df.AVG.str.replace(",", "")
df.AVG = df.AVG.astype(float)
df.plot(kind='scatter', x="GRADE", y="AVG") # , 'rx')
plt.show()