def test_simulate():
tree = MonteCarlo(player)
tree.make_node(initial_state, player)
key = hash(initial_state.tostring()) + hash(player)
root = tree.nodes[key]
for _ in range(1000):
outcome = tree.simulate(root)
assert isinstance(outcome, BoardPiece) or isinstance(
outcome, GameState)
def test_backpropagate():
tree = MonteCarlo(player)
tree.make_node(initial_state, player)
key = hash(initial_state.tostring()) + hash(player)
root = tree.nodes[key]
n_test_sims = 100
for m in root.legal_moves:
wins = 0
child = tree.expand(root)
for i in range(n_test_sims):
outcome = tree.simulate(child)
tree.backpropagate(child, outcome)
if outcome == player:
wins += 1
assert child.n_wins == wins
assert child.n_plays == n_test_sims
assert root.n_plays == sum([c.n_plays for c in root.children.values()])
assert root.n_wins == sum([c.n_wins for c in root.children.values()])