def getNumFreq(lead): inLead=lt.inLead(lead) changes=lt.numChanges(inLead, lead) print np.mean(changes) h,b=np.histogram(changes, max(changes)-min(changes)) hcorr=h/float(len(changes)) return hcorr,b
def getLastLeadFreq(lead, bins):
inLead = lt.inLead(lead)
s = lt.lastChange(inLead)
h, b = np.histogram(s, bins)
binw = arcs.getBinWidth(bins)
hcorr = h / (binw * len(s))
return hcorr, b
def getLastLeadFreq(lead, bins): inLead=lt.inLead(lead) s=lt.lastChange(inLead) h,b=np.histogram(s, bins) binw=arcs.getBinWidth(bins) hcorr=h/(binw*len(s)) return hcorr, b
def getNumFreq(lead):
inLead = lt.inLead(lead)
changes = lt.numChanges(inLead, lead)
print np.mean(changes)
h, b = np.histogram(changes, max(changes) - min(changes))
hcorr = h / float(len(changes))
return hcorr, b
def plotBathtub(sport):
if sport=='NBA':
scope=2880
seasons='2002-2010'
bins=NBAbins()
else:
scope=3600
seasons='2000-2009'
if sport=='NHL':
bins=NHLbins()
elif sport=='CFB':
bins=NFLbins()
elif sport=='NFL':
bins=NFLbins()
binw=getBinWidth(bins)
step=120.0
data=lt.getData(sport)
lead=lt.Lead(data, sport)
inLead=lt.inLead(lead)
s=lt.lastChange(inLead)
#m,s=lt.maxLeadTime(lead)
'''rw_lead=rwk.Lead(sport)
rw_inLead=rwk.inLead(rw_lead)
rw=rwk.lastChange(rw_inLead)'''
sm_lead=sm.Lead(sport)
sm_inLead=sm.inLead(sm_lead)
sr=sm.lastChange(sm_inLead)
msr,sr=lt.maxLeadTime(sm_lead)
fontSize=18
ax=plt.gca()
ax.tick_params(labelsize=fontSize)
h,b=np.histogram(s, bins)
hcorr=h/(binw*len(s))
plt.scatter(b[:len(bins)-1], hcorr, c='blue',marker='o',label=sport+' games')
hr,br=np.histogram(sr,bins)
hrcorr=hr/(binw*len(sr))
plt.plot(br[:len(bins)-1],hrcorr,color='DarkTurquoise',linewidth=2,label='Inhomogeneous Poisson process')
#ubiased RW
'''rwstep=20.0
rwbins=scope/rwstep
h,b=np.histogram(rw, rwbins)
hcorr=h/(rwstep*len(rw))
plt.plot(b[:rwbins], hcorr, c='purple',linewidth=2.5,label='Homogenous Poisson process') '''
x=np.array(range(scope)) #arcsine law
y=1/(np.pi*(x*(scope+1-x))**(0.5))
plt.plot(x,y,color='FireBrick',linewidth=2,label='Arcsine law')
plt.xlim(xmin=0,xmax=scope)
plt.ylim(ymin=0,ymax=0.0027)
plt.legend(prop={'size':fontSize})
plt.xlabel('Game clock time, t (seconds)', fontsize=fontSize)
plt.ylabel('Probability of maximum lead change', fontsize=fontSize)
plt.subplots_adjust(left=0.16, right=0.95, top=0.95, bottom=0.13)
#plt.savefig(sport+'varBinnedUpdate.pdf')
#plt.close()
plt.show()
def getSafeAvg(lead, sport): inLead=lt.inLead(lead) lastChange=lt.lastChangeMessy(inLead) safe=lt.assignSafety(lastChange, sport) z_tuples=lt.getZ(lead, safe, sport) df=pd.DataFrame(z_tuples) z1,ave_q=binAverage(df) return z1,ave_q
def getSafeAvg(lead, sport):
inLead = lt.inLead(lead)
lastChange = lt.lastChangeMessy(inLead)
safe = lt.assignSafety(lastChange, sport)
z_tuples = lt.getZ(lead, safe, sport)
df = pd.DataFrame(z_tuples)
z1, ave_q = binAverage(df)
return z1, ave_q
def plotWScoringRate(sport):
if sport=='NBA':
scope=2880
seasons='2002-2010'
bins=arcs.NBAbins()
else:
scope=3600
seasons='2000-2009'
if sport=='NHL':
bins=arcs.NHLbins()
elif sport=='CFB':
bins=arcs.NFLbins()
elif sport=='NFL':
bins=arcs.NFLbins()
binw=arcs.getBinWidth(bins)
step=120.0
data=lt.getData(sport)
lead=lt.Lead(data, sport)
inLead=lt.inLead(lead)
s=lt.lastChange(inLead)
#m,s=lt.maxLeadTime(lead)
'''rw_lead=rwk.Lead(sport)
rw_inLead=rwk.inLead(rw_lead)
rw=rwk.lastChange(rw_inLead)'''
sm_lead=sm.Lead(sport)
sm_inLead=sm.inLead(sm_lead)
sr=sm.lastChange(sm_inLead)
#msr,sr=lt.maxLeadTime(sm_lead)
ev_prob=pd.DataFrame.from_csv('/Users/Ish/Documents/SafeLeads/Results/'+sport+'_res/'+sport+'_eventProb.csv',\
header=None)
fontSize=18
window=10
#f, (ax1, ax2) = plt.subplots(2, sharex=True)
f=plt.figure()
gs=gridspec.GridSpec(2,1,height_ratios=[2,3]) #1 used to be 3
ax1=plt.subplot(gs[0,:])
ax2=plt.subplot(gs[1:,:], sharex=ax1)
smoothEvProb=movingaverage(ev_prob[1],window)
avg=np.mean(smoothEvProb)
#y_formatter = matplotlib.ticker.ScalarFormatter(useOffset=-100)
#ax1.yaxis.set_major_formatter(y_formatter)
ax1.plot(smoothEvProb, linewidth=1.3)
ax1.hlines(avg, 0,scope, colors='red', linewidth=2)
yticks = ax1.yaxis.get_major_ticks()
yticks[0].label1.set_visible(False)
ax1.set_ylim(ymin=0,ymax=max(smoothEvProb)+0.005)
ax1.set_ylabel('Pr(scoring event)',fontsize=fontSize, labelpad=25)#25 for NBA
ax1.tick_params(labelsize=fontSize)
ax2.tick_params(labelsize=fontSize)
h,b=np.histogram(s, bins)
hcorr=h/(binw*len(s))
ax2.scatter(b[:len(bins)-1], hcorr, c='blue',marker='o',label=sport+' games')
hr,br=np.histogram(sr,bins)
hrcorr=hr/(binw*len(sr))
ax2.plot(br[:len(bins)-1],hrcorr,color='DarkTurquoise',linewidth=2,label='Inhomogeneous Poisson process')
#ubiased RW
'''rwstep=20.0
rwbins=scope/rwstep
h,b=np.histogram(rw, rwbins)
hcorr=h/(rwstep*len(rw))
plt.plot(b[:rwbins], hcorr, c='purple',linewidth=2.5,label='Homogenous Poisson process') '''
x=np.array(range(scope)) #arcsine law
y=1/(np.pi*(x*(scope+1-x))**(0.5))
ax2.plot(x,y,color='FireBrick',linewidth=2,label='Arcsine law')
ax2.set_xlim(xmin=0,xmax=scope)
ax2.set_ylim(ymin=0,ymax=0.0027)
ax2.legend(prop={'size':fontSize})
ax2.set_xlabel('Game clock time, t (seconds)', fontsize=fontSize)
ax2.set_ylabel('Pr(last lead change)', fontsize=fontSize)
f.subplots_adjust(left=0.16, right=0.95, top=0.95, bottom=0.11, hspace=0.00001) #16,13
# Fine-tune figure; make subplots close to each other and hide x ticks for
# all but bottom plot.
plt.setp([a.get_xticklabels() for a in f.axes[:-1]], visible=False)
plt.show()
def PropInLead(sport):
#portion of time first team to score is in the lead
data=lt.getData(sport)
lead=lt.Lead(data,sport)
inLead=lt.inLead(lead)
rands=np.random.rand(len(inLead))
mult=-1+2*(rands<0.5)
matr=(np.tile(mult,(len(inLead[0]),1))).T
rand_inLead=np.multiply(matr, inLead)
props=np.sum(rand_inLead<0, axis=1)
sm_lead=sm.Lead(sport)
sm_inLead=sm.inLead(sm_lead)
rands=np.random.rand(len(sm_inLead))
mult=-1+2*(rands<0.5)
matr=(np.tile(mult,(len(sm_inLead[0]),1))).T
sm_randInLead=np.multiply(matr,sm_inLead)
sm_props=np.sum(sm_randInLead<0,axis=1)
rw_lead=rw.Lead(sport)
rw_inLead=rw.inLead(rw_lead)
rands=np.random.rand(len(rw_inLead))
mult=-1+2*(rands<0.5)
matr=(np.tile(mult,(len(rw_inLead[0]),1))).T
rw_randInLead=np.multiply(matr, rw_inLead)
rw_props=np.sum(rw_randInLead<0,axis=1)
if sport=='NBA':
scope=2880
bins=arcs.NBAbins()
else:
scope=3600
if sport=='NHL':
bins=arcs.NHLbins()
else:
bins=arcs.NFLbins()
binw=arcs.getBinWidth(bins)
fontSize=18
ax=plt.gca()
ax.tick_params(labelsize=fontSize)
h,b=np.histogram(props, bins)
hcorr=h/(binw*len(props))
plt.scatter(b[:len(bins)-1], hcorr, c='blue',marker='o',label=sport+' games')
hr,br=np.histogram(sm_props,bins)
hrcorr=hr/(binw*len(sm_props))
plt.plot(br[:len(bins)-1],hrcorr,color='DarkTurquoise', linewidth=2,label='Inhomogeneous Poisson process')
#Unbiased rw
#rwstep=20.0
#rwbins=scope/rwstep
#hw,bw=np.histogram(rw_props,rwbins)
#hwcorr=hw/(rwstep*len(rw_props))
#plt.plot(bw[:rwbins],hwcorr,color='DarkSalmon', linewidth=2,label='Homogeneous Poisson process')
#arcsine law
x=np.array(range(scope))
y=1/(np.pi*(x*(scope+1-x))**(0.5))
plt.plot(x,y,color='FireBrick',linewidth=2,label='Arcsine law')
plt.xlim(xmin=0,xmax=scope)
plt.ylim(ymin=0,ymax=0.0027)
plt.legend(prop={'size':fontSize})
plt.xlabel('Number of seconds a team is in the lead', fontsize=fontSize)
plt.ylabel('Relative frequency', fontsize=fontSize)
plt.subplots_adjust(left=0.16, right=0.95, top=0.95, bottom=0.13)
#plt.savefig(sport+'varBinnedUpdate.pdf')
#plt.close()
plt.show()
def plotWScoringRate(sport):
if sport == 'NBA':
scope = 2880
seasons = '2002-2010'
bins = arcs.NBAbins()
else:
scope = 3600
seasons = '2000-2009'
if sport == 'NHL':
bins = arcs.NHLbins()
elif sport == 'CFB':
bins = arcs.NFLbins()
elif sport == 'NFL':
bins = arcs.NFLbins()
binw = arcs.getBinWidth(bins)
step = 120.0
data = lt.getData(sport)
lead = lt.Lead(data, sport)
inLead = lt.inLead(lead)
s = lt.lastChange(inLead)
#m,s=lt.maxLeadTime(lead)
'''rw_lead=rwk.Lead(sport)
rw_inLead=rwk.inLead(rw_lead)
rw=rwk.lastChange(rw_inLead)'''
sm_lead = sm.Lead(sport)
sm_inLead = sm.inLead(sm_lead)
sr = sm.lastChange(sm_inLead)
#msr,sr=lt.maxLeadTime(sm_lead)
ev_prob=pd.DataFrame.from_csv('/Users/Ish/Documents/SafeLeads/Results/'+sport+'_res/'+sport+'_eventProb.csv',\
header=None)
fontSize = 18
window = 10
#f, (ax1, ax2) = plt.subplots(2, sharex=True)
f = plt.figure()
gs = gridspec.GridSpec(2, 1, height_ratios=[2, 3]) #1 used to be 3
ax1 = plt.subplot(gs[0, :])
ax2 = plt.subplot(gs[1:, :], sharex=ax1)
smoothEvProb = movingaverage(ev_prob[1], window)
avg = np.mean(smoothEvProb)
#y_formatter = matplotlib.ticker.ScalarFormatter(useOffset=-100)
#ax1.yaxis.set_major_formatter(y_formatter)
ax1.plot(smoothEvProb, linewidth=1.3)
ax1.hlines(avg, 0, scope, colors='red', linewidth=2)
yticks = ax1.yaxis.get_major_ticks()
yticks[0].label1.set_visible(False)
ax1.set_ylim(ymin=0, ymax=max(smoothEvProb) + 0.005)
ax1.set_ylabel('Pr(scoring event)', fontsize=fontSize,
labelpad=25) #25 for NBA
ax1.tick_params(labelsize=fontSize)
ax2.tick_params(labelsize=fontSize)
h, b = np.histogram(s, bins)
hcorr = h / (binw * len(s))
ax2.scatter(b[:len(bins) - 1],
hcorr,
c='blue',
marker='o',
label=sport + ' games')
hr, br = np.histogram(sr, bins)
hrcorr = hr / (binw * len(sr))
ax2.plot(br[:len(bins) - 1],
hrcorr,
color='DarkTurquoise',
linewidth=2,
label='Inhomogeneous Poisson process')
#ubiased RW
'''rwstep=20.0
rwbins=scope/rwstep
h,b=np.histogram(rw, rwbins)
hcorr=h/(rwstep*len(rw))
plt.plot(b[:rwbins], hcorr, c='purple',linewidth=2.5,label='Homogenous Poisson process') '''
x = np.array(range(scope)) #arcsine law
y = 1 / (np.pi * (x * (scope + 1 - x))**(0.5))
ax2.plot(x, y, color='FireBrick', linewidth=2, label='Arcsine law')
ax2.set_xlim(xmin=0, xmax=scope)
ax2.set_ylim(ymin=0, ymax=0.0027)
ax2.legend(prop={'size': fontSize})
ax2.set_xlabel('Game clock time, t (seconds)', fontsize=fontSize)
ax2.set_ylabel('Pr(last lead change)', fontsize=fontSize)
f.subplots_adjust(left=0.16,
right=0.95,
top=0.95,
bottom=0.11,
hspace=0.00001) #16,13
# Fine-tune figure; make subplots close to each other and hide x ticks for
# all but bottom plot.
plt.setp([a.get_xticklabels() for a in f.axes[:-1]], visible=False)
plt.show()
def PropInLead(sport):
#portion of time first team to score is in the lead
data = lt.getData(sport)
lead = lt.Lead(data, sport)
inLead = lt.inLead(lead)
rands = np.random.rand(len(inLead))
mult = -1 + 2 * (rands < 0.5)
matr = (np.tile(mult, (len(inLead[0]), 1))).T
rand_inLead = np.multiply(matr, inLead)
props = np.sum(rand_inLead < 0, axis=1)
sm_lead = sm.Lead(sport)
sm_inLead = sm.inLead(sm_lead)
rands = np.random.rand(len(sm_inLead))
mult = -1 + 2 * (rands < 0.5)
matr = (np.tile(mult, (len(sm_inLead[0]), 1))).T
sm_randInLead = np.multiply(matr, sm_inLead)
sm_props = np.sum(sm_randInLead < 0, axis=1)
rw_lead = rw.Lead(sport)
rw_inLead = rw.inLead(rw_lead)
rands = np.random.rand(len(rw_inLead))
mult = -1 + 2 * (rands < 0.5)
matr = (np.tile(mult, (len(rw_inLead[0]), 1))).T
rw_randInLead = np.multiply(matr, rw_inLead)
rw_props = np.sum(rw_randInLead < 0, axis=1)
if sport == 'NBA':
scope = 2880
bins = arcs.NBAbins()
else:
scope = 3600
if sport == 'NHL':
bins = arcs.NHLbins()
else:
bins = arcs.NFLbins()
binw = arcs.getBinWidth(bins)
fontSize = 18
ax = plt.gca()
ax.tick_params(labelsize=fontSize)
h, b = np.histogram(props, bins)
hcorr = h / (binw * len(props))
plt.scatter(b[:len(bins) - 1],
hcorr,
c='blue',
marker='o',
label=sport + ' games')
hr, br = np.histogram(sm_props, bins)
hrcorr = hr / (binw * len(sm_props))
plt.plot(br[:len(bins) - 1],
hrcorr,
color='DarkTurquoise',
linewidth=2,
label='Inhomogeneous Poisson process')
#Unbiased rw
#rwstep=20.0
#rwbins=scope/rwstep
#hw,bw=np.histogram(rw_props,rwbins)
#hwcorr=hw/(rwstep*len(rw_props))
#plt.plot(bw[:rwbins],hwcorr,color='DarkSalmon', linewidth=2,label='Homogeneous Poisson process')
#arcsine law
x = np.array(range(scope))
y = 1 / (np.pi * (x * (scope + 1 - x))**(0.5))
plt.plot(x, y, color='FireBrick', linewidth=2, label='Arcsine law')
plt.xlim(xmin=0, xmax=scope)
plt.ylim(ymin=0, ymax=0.0027)
plt.legend(prop={'size': fontSize})
plt.xlabel('Number of seconds a team is in the lead', fontsize=fontSize)
plt.ylabel('Relative frequency', fontsize=fontSize)
plt.subplots_adjust(left=0.16, right=0.95, top=0.95, bottom=0.13)
#plt.savefig(sport+'varBinnedUpdate.pdf')
#plt.close()
plt.show()
def plotBathtub(sport):
if sport == 'NBA':
scope = 2880
seasons = '2002-2010'
bins = NBAbins()
else:
scope = 3600
seasons = '2000-2009'
if sport == 'NHL':
bins = NHLbins()
elif sport == 'CFB':
bins = NFLbins()
elif sport == 'NFL':
bins = NFLbins()
binw = getBinWidth(bins)
step = 120.0
data = lt.getData(sport)
lead = lt.Lead(data, sport)
inLead = lt.inLead(lead)
s = lt.lastChange(inLead)
#m,s=lt.maxLeadTime(lead)
'''rw_lead=rwk.Lead(sport)
rw_inLead=rwk.inLead(rw_lead)
rw=rwk.lastChange(rw_inLead)'''
sm_lead = sm.Lead(sport)
sm_inLead = sm.inLead(sm_lead)
sr = sm.lastChange(sm_inLead)
msr, sr = lt.maxLeadTime(sm_lead)
fontSize = 18
ax = plt.gca()
ax.tick_params(labelsize=fontSize)
h, b = np.histogram(s, bins)
hcorr = h / (binw * len(s))
plt.scatter(b[:len(bins) - 1],
hcorr,
c='blue',
marker='o',
label=sport + ' games')
hr, br = np.histogram(sr, bins)
hrcorr = hr / (binw * len(sr))
plt.plot(br[:len(bins) - 1],
hrcorr,
color='DarkTurquoise',
linewidth=2,
label='Inhomogeneous Poisson process')
#ubiased RW
'''rwstep=20.0
rwbins=scope/rwstep
h,b=np.histogram(rw, rwbins)
hcorr=h/(rwstep*len(rw))
plt.plot(b[:rwbins], hcorr, c='purple',linewidth=2.5,label='Homogenous Poisson process') '''
x = np.array(range(scope)) #arcsine law
y = 1 / (np.pi * (x * (scope + 1 - x))**(0.5))
plt.plot(x, y, color='FireBrick', linewidth=2, label='Arcsine law')
plt.xlim(xmin=0, xmax=scope)
plt.ylim(ymin=0, ymax=0.0027)
plt.legend(prop={'size': fontSize})
plt.xlabel('Game clock time, t (seconds)', fontsize=fontSize)
plt.ylabel('Probability of maximum lead change', fontsize=fontSize)
plt.subplots_adjust(left=0.16, right=0.95, top=0.95, bottom=0.13)
#plt.savefig(sport+'varBinnedUpdate.pdf')
#plt.close()
plt.show()