Author | Sean |
Submission date | 2016-02-10 11:26:21.773619 |
Rating | 6806 |
Matches played | 397 |
Win rate | 68.01 |
Use rpsrunner.py to play unranked matches on your computer.
if input == "":
import random
class MarkovChain:
def __init__(self, counts = None):
self.visits = 0
if counts is None:
self.counts = [0 for _ in xrange(3)]
else:
self.counts = counts
self.children = None
def split_edge(self, i):
old = self.children[i]
new = MarkovChain(old.counts)
self.children[i] = new
new.children = old.children
def transition(self, i, j, t):
self.visits += 1
self.counts[i] += 1
if t == 0:
k = 3 * i + j
elif t == 1:
k = i
else:
k = j
if self.children[k].visits >= 2:
self.split_edge(k)
return self.children[k]
R, P, S = 0, 1, 2
index = {"R": R, "P": P, "S": S}
beat = ("P", "S", "R")
l = 3
def new_model(n):
nodes = [[MarkovChain() for _ in xrange(n)] for _ in xrange(l)]
for i in xrange(l):
children = nodes[(i + 1) % l]
for j in xrange(n):
nodes[i][j].children = children
model = MarkovChain()
model.children = nodes[0]
return model
both = new_model(9)
them = new_model(3)
us = new_model(3)
else:
i = index[input]
j = index[output]
both = both.transition(i, j, 0)
them = them.transition(i, j, 1)
us = us.transition(i, j, 2)
counts = [a + b + c + 1 for a, b, c in zip(both.counts,
them.counts,
us.counts)]
t = sum(counts)
r = random.randint(0, t)
x = 0
for i, p in enumerate(counts):
x += p
if r <= x:
output = beat[i]
break