def test_get_wins_loses(self):
diff_teams_pairs = {(1, 4): [1, 2]}
otpc = DifferentTeamsPairsContainer(diff_teams_pairs)
self.assertEqual(otpc.get_wins_loses(1, 4), (1, 2))
self.assertEqual(otpc.get_wins_loses(4, 1), (2, 1))
self.assertEqual(otpc.get_wins_loses(5, 7), (0, 0))
def test_get_wins_loses(self):
diff_teams_pairs = {(1, 4): [1, 2]}
otpc = DifferentTeamsPairsContainer(diff_teams_pairs)
self.assertEqual(otpc.get_wins_loses(1, 4), (1, 2))
self.assertEqual(otpc.get_wins_loses(4, 1), (2, 1))
self.assertEqual(otpc.get_wins_loses(5, 7), (0, 0))
hero_id_to_name = get_hero_id_to_name()
def show_win_loses(heroes, hr_id_to_nm):
if isinstance(heroes, dict):
heroes = heroes.items()
for key, value in heroes:
if isinstance(key, int):
key = (key, )
names = ["{0:20}".format(hero_id_to_name[id]) for id in key]
name = "| ".join(names)
print "{0} {1:0>5} {2:0>5}".format(name, *value)
solo_hr = defaultdict(lambda: [0, 0])
stpc = SameTeamsPairsContainer()
dtpc = DifferentTeamsPairsContainer()
WIN_INDEX, LOSE_INDEX = 0, 1
same_triple = defaultdict(lambda: [0, 0])
for i, record in enumerate(train_data):
radiant_win, radiant, dire = record
winners, losers = (radiant, dire) if radiant_win else (dire, radiant)
#solo heroes
for hero_id in winners:
solo_hr[hero_id][WIN_INDEX] += 1
for hero_id in losers:
solo_hr[hero_id][LOSE_INDEX] += 1
for hero1_id, hero2_id in combinations(winners, 2):
STAT_FILE_NAME = "stat_data.p"
TEST_DATA_FILE_NAME = "test_data16451.p"
TEAM_OFFSET = 1000
print "load stat data from {0}".format(STAT_FILE_NAME)
stat_data = pickle.load(open(STAT_FILE_NAME, 'rb'))
print "load complete"
print "load test data from {0}".format(TEST_DATA_FILE_NAME)
test_data = pickle.load(open(TEST_DATA_FILE_NAME, 'rb'))
print "load complete"
solo_hr, stpc, dtpc = stat_data
solo_hr = defaultdict(lambda: [0, 0], solo_hr)
stpc = SameTeamsPairsContainer(stpc)
dtpc = DifferentTeamsPairsContainer(dtpc)
tr = FractionTermResolver(TEAM_OFFSET, solo_hr, stpc, dtpc, smoothing=0.5)
#tr = OddsTermResolver(TEAM_OFFSET, solo_hr, stpc, dtpc)
predictor = PredictorNaiveBayesBigram(TEAM_OFFSET, tr)
num_correct = 0.0
num_matches = len(test_data)
for record in test_data:
radiant_win, radiant, dire = record
probability = predictor.predict(radiant, dire)
if probability > 0.5001 and radiant_win or probability < 0.4999 and not radiant_win:
num_correct += 1.0
def show_win_loses(heroes, hr_id_to_nm):
if isinstance(heroes, dict):
heroes = heroes.items()
for key, value in heroes:
if isinstance(key, int):
key = (key, )
names = ["{0:20}".format(hero_id_to_name[id]) for id in key]
name = "| ".join(names)
print "{0} {1:0>5} {2:0>5}".format(name, *value)
solo_hr = defaultdict(lambda: [0, 0])
stpc = SameTeamsPairsContainer()
dtpc = DifferentTeamsPairsContainer()
WIN_INDEX, LOSE_INDEX = 0, 1
same_triple = defaultdict(lambda: [0, 0])
for i, record in enumerate(train_data):
radiant_win, radiant, dire = record
winners, losers = (radiant, dire) if radiant_win else (dire, radiant)
#solo heroes
for hero_id in winners:
solo_hr[hero_id][WIN_INDEX] += 1
for hero_id in losers:
solo_hr[hero_id][LOSE_INDEX] += 1
for hero1_id, hero2_id in combinations(winners, 2):
def setUp(self):
self.otpc = DifferentTeamsPairsContainer()