def make_hist():
"""
Makes a histogram of the at-home ppg data.
The histogram contains lines for the at-home ppg average and
for the total ppg average.
"""
ppgs = get_data('data/home_ppg.dat')
average = np.mean(ppgs)
totalppgs = get_data('data/total_ppg.dat')
totalppg_avg = np.mean(totalppgs)
f, (ax1, ax2) = plt.subplots(2, sharex=True)
ax1.hist(totalppgs, bins=12, color='gray')
ax1.plot([totalppg_avg, totalppg_avg], [0, 700], lw=3, color='g')
ax1.text(
totalppg_avg+1, 620, "total mean=%1.3f" % totalppg_avg, size='large')
ax1.text(-3, 520, "all games", rotation=90, size='large')
ax1.set_yticks([])
ax2.hist(ppgs, bins=12, color='gray')
ax2.plot([average, average], [0, 350], lw=3, color='b')
ax2.text(average+1, 300, "home mean=%1.3f" % average, size='large')
ax2.text(-3, 270, "home games", rotation=90, size='large')
plt.yticks([])
plt.xlabel('points per game', size='large')
plt.show()
def plot_data(years, stat, win_type):
data = get_data('data/%s_top_%s.dat' % (years, stat))
plt.figure(1)
min_x, max_x = sys.maxint, 0
min_y, max_y = sys.maxint, 0
for datum in data:
name, value, passes_filter, home_wins, total_wins = datum
if win_type == 'home':
wins = home_wins
elif win_type == 'total':
wins = total_wins
else:
print "Warning: invalid win_type. Defaulted to 'home'."
wins = home_wins
if passes_filter:
color = 'k'
else:
color = 'r'
plt.text(wins, value, name, size='large', color=color)
min_x, max_x = min(min_x, wins), max(max_x, wins)
min_y, max_y = min(min_y, value), max(max_y, value)
plt.xlim(min_x - max_x/10.0, max_x + max_x/10.0)
plt.ylim(min_y - max_y/10.0, max_y + max_y/10.0)
# X and Y labels
plt.ylabel(stat, size='large')
plt.xlabel('# of %s wins' % win_type, size='large')
plt.show()
def make_boxplot():
"""
Makes three boxplots for home, away, and total ppgs.
"""
home = get_data('data/home_ppg.dat')
away = get_data('data/away_ppg.dat')
total = get_data('data/total_ppg.dat')
ax = plt.subplot(111)
plt.boxplot([home, away, total], vert=False, widths=0.8)
plt.yticks([])
plt.text(-3, 3.3, "all games", rotation=90, size='large')
plt.text(-3, 2.3, "away games", rotation=90, size='large')
plt.text(-3, 1.3, "home games", rotation=90, size='large')
plt.xlabel("points per game")
plt.xlim(-1, 70)
ax.xaxis.label.set_fontsize(20)
plt.show()
def plot_data_comparison(years_x, stat_x, years_y, stat_y, square):
data_x = get_data('data/%s_top_%s.dat' % (years_x, stat_x))
data_y = get_data('data/%s_top_%s.dat' % (years_y, stat_y))
plt.figure(1)
min_x, max_x = sys.maxint, 0
min_y, max_y = sys.maxint, 0
data = {}
# Links a team with their stat_x.
for datum in data_x:
name, value = datum[0], datum[1]
data[name] = [value]
# Links a team with their stat_y. for datum in data_y:
for datum in data_y:
name, value = datum[0], datum[1]
data[name].append(value)
# Plot
for team in data:
x_val, y_val = data[team]
plt.text(x_val, y_val, team, size='large')
min_x, max_x = min(min_x, x_val), max(max_x, x_val)
min_y, max_y = min(min_y, y_val), max(max_y, y_val)
if square:
min_x, min_y = min(min_x, min_y), min(min_x, min_y)
max_x, max_y = max(max_x, max_y), max(max_x, max_y)
x_diff = max_x - min_x
y_diff = max_y - min_y
plt.xlim(min_x - x_diff/6.0, max_x + x_diff/6.0)
plt.ylim(min_y - y_diff/6.0, max_y + y_diff/6.0)
# X and Y labels
plt.ylabel(stat_y, size='large')
plt.xlabel(stat_x, size='large')
plt.show()
def plot_box_and_hist(filename, xlabel, comparison_val):
"""
Makes a box plot and histogram of the data contained in the filename.
Also draws in a line for a comparision value.
For example, if plotting the at-home point spreads, a good comparison
value would be 0.0.
"""
data = get_data(filename)
average = np.mean(data)
fig = plt.figure(figsize=(6, 4))
boxplot_axes = fig.add_axes([0.1, 0.7, 0.7, 0.15])
histplot_axes = fig.add_axes([0.1, 0.1, 0.7, 0.6])
histplot_axes.xaxis.label.set_fontsize(13)
histplot_axes.yaxis.label.set_fontsize(13)
for item in (histplot_axes.get_xticklabels() +
histplot_axes.get_yticklabels()):
item.set_fontsize(13)
boxplot_axes.boxplot(data, notch=True, vert=False)
boxplot_axes.plot([average, average], [0, 3], lw=3, color='g')
boxplot_axes.plot(
[comparison_val, comparison_val], [0, 3], lw=3, color='b')
hist = histplot_axes.hist(data, bins=12, color='gray')
maxval = max(hist[0])
histplot_axes.plot(
[average, average],
[0, maxval+maxval/6],
lw=3,
color='g'
)
histplot_axes.plot(
[comparison_val, comparison_val],
[0, maxval+maxval/6],
lw=3,
color='b'
)
histplot_axes.set_ylim(0, maxval+maxval/6)
boxplot_axes.set_xticklabels([])
boxplot_axes.set_yticks([])
histplot_axes.set_yticks([])
plt.xlabel(xlabel, size='large')
plt.show()
Both tests result with very large t-values, mainly because the sample size
is so large (over 1500 games). I don't find this very interesting.
"""
def get_std(dat1, dat2):
mean1 = np.mean(dat1)
mean2 = np.mean(dat2)
summed = (sum((x - mean1)**2 for x in dat1) +
sum((x - mean2)**2 for x in dat2))
return summed / (len(dat1) + len(dat2) - 2.0)
if __name__ == "__main__":
home = get_data('data/home_ppg.dat')
home_mean, home_n = np.mean(home), len(home)
away = get_data('data/away_ppg.dat')
away_mean, away_n = np.mean(away), len(away)
std = get_std(home, away)
print (home_mean - away_mean) / sqrt(std * (1.0/home_n + 1.0/away_n))
spread = []
for i in xrange(len(home)):
spread.append(home[i] - away[i])
spread_mean = np.mean(spread)
spread_std = np.std(spread)
f = file('data/home_spread.dat', 'w')
for val in spread:
