def getprob(self, item1, item2):
import choix
from web.util import nanguard
a = self.getid(item1["hash"] if type(item1) == dict else item1)
b = self.getid(item2["hash"] if type(item2) == dict else item2)
a_new = False
b_new = False
if a >= len(self.model):
a_new = True
if b >= len(self.model):
b_new = True
#self.calculate_ranking()
if not len(self.model) or a_new or b_new:
return 0.5, 0.5
#return f"no model yet. new: {a_new}", f"no model yet. new: {b_new}"
ra, rb = choix.probabilities([a, b], self.model)
return nanguard(ra, "ra"), nanguard(rb, "rb")
def pick_next_comparison(data, n_items):
params = choix.ilsr_pairwise(n_items, data, alpha=0.01)
count = 10
while count > 0:
x = random.randint(0, n_items - 1)
y = random.randint(0, n_items - 1)
while y == x:
y = random.randint(0, n_items - 1)
prob_x_wins, prob_y_wins = choix.probabilities([x, y], params)
if abs(
0.5 - prob_x_wins
) < 0.1: #if comparison has below 60% chance of being guessed correctly atm
return x, y
count -= 1
#if search 10 times for good comparison and fail pick random one
x = random.randint(0, n_items - 1)
y = random.randint(0, n_items - 1)
while y == x:
y = random.randint(0, n_items - 1)
return x, y
namps = amps.size #number of amplitudes
BTdata = []
for i in range(len(data)): #Go though each result file
for ii in range(np.size(data[i], axis=0)):
if data[i, ii, 3] == 1.0:
BTdata.append((np.int(
(data[i, ii, 2]) * 2), np.int((data[i, ii, 1]) * 2)))
else:
BTdata.append((np.int(
(data[i, ii, 1]) * 2), np.int((data[i, ii, 2]) * 2)))
BT = choix.ilsr_pairwise(namps, BTdata, 0.1)
pi = choix.probabilities(list(range(namps)), BT)
pyplot.close('all')
dx = 0.02
textX = 0.6 + dx
textY = 0.13
ax = pyplot.axes()
tsize = 12
pyplot.plot(Xdata, pi, 'o-', color='black')
pyplot.axvline(0.63, linestyle='--', color='gray')
ax.text(
textX,
textY,
'Human kinematic',
import pandas as pd
import numpy as np
import choix
df = pd.read_csv("nba1617.csv")
teams = list(set(df.Home.unique()))
t = list(np.sort(teams))
data = []
for i in range(len(df.Visitor)):
if df.PTSV[i] > df.PTSH[i]:
el = (t.index(df.Visitor[i]), t.index(df.Home[i]))
else:
el = (t.index(df.Home[i]), t.index(df.Visitor[i]))
data.append(el)
skills = choix.ilsr_pairwise(30, data)
for i in range(30):
win_probs = []
for j in range(30):
win_probs.append(choix.probabilities([i, j], skills)[0])
print win_probs