cfr.py

from Node import Node

def cfr(nodes, game, gameState, p0, p1):
	# Player determination
	player = game.getCurrentPlayer(gameState)
		
	# game events
	gameState = game.applyGameEvents(gameState)
		
	# Detect terminal game nodes and return earnings 
	if game.isTerminal(gameState):
		opponent = 1 - player
		return  game.getOutcome(gameState, player, opponent)
			
	# get or create current game node
	infoSet = game.getInfoSet(gameState)
	node = None
	if not infoSet in nodes:
		node = game.createNode(gameState)
		nodes[infoSet] = node
	else:
		node = nodes[infoSet]
	
	# calculate hypothetical returns for all actions
	realizationWeight = [p0, p1][player]
	strategy = node.getStrategy(realizationWeight)
	util = [0] * node.getActionCount()
	nodeUtil = 0
	for a in range(node.getActionCount()):
		newState = gameState.nextHistory(a)
		# negate opponent's regrets
		util[a] = -1.0 * cfr(nodes, game, newState, [p0 * strategy[a], p0][player], [p1, p1 * strategy[a]][player])
		nodeUtil += strategy[a] * util[a]

	# update regrets
	for a in range(node.getActionCount()):
		regret = util[a] - nodeUtil
		node.regretSum[a] += [p1, p0][player] * regret

	return nodeUtil