feat: New method in play_vs_alphazero. We can now let the alphazero-a…

…lgorithm play against a standard MCTS algorithm. Sadly, alphazero is still not able to win against the standard MCTS algorithm when both are given 1000 simulations.

src/play/play_vs_alphazero.py

@@ -1,5 +1,5 @@
-from src.neuralnet.neural_network import NeuralNetwork
 from src.alphazero.agents.alphazero_play_agent import AlphaZero
+from src.mcts.mcts import Mcts
 from src.utils.game_context import GameContext
 
 def player(state):
@@ -17,11 +17,38 @@ def player(state):
             print("Invalid move, try again!")
     state.apply_action(action)
 
-
-def ai(state, alphazero: AlphaZero):
+def ai(state, alphazero: AlphaZero | Mcts):
     action = alphazero.run_simulation(state, num_simulations = 1000)
     state.apply_action(action)
 
+def play_mcts_vs_alphazero(context: GameContext, first: bool):
+    """
+    Lets the MCTS agent play against the AlphaZero agent.
+    If first is True, the MCTS agent will play first.
+    Both agents are given 1000 simulations before making a move.
+    """
+
+    mcts = Mcts()
+    alphazero = AlphaZero(context=context)
+    state = alphazero.context.game.new_initial_state()
+
+    if not first:
+        print('~~~~~~~~~~~~~~~ ALPHAZERO ~~~~~~~~~~~~~~~~')
+        print(state)
+        ai(state, alphazero)
+
+    while not state.is_terminal():
+        print('~~~~~~~~~~~~~~~ MONTE CARLO ~~~~~~~~~~~~~~~~')
+        print(state)
+        ai(state, mcts)
+        print('~~~~~~~~~~~~~~~ ALPHAZERO ~~~~~~~~~~~~~~~~')
+        print(state)
+        if state.is_terminal():
+            break
+        ai(state, alphazero)
+        if state.is_terminal():
+            print('~~~~~~~~~~~~~~~ MONTE CARLO ~~~~~~~~~~~~~~~~')
+            print(state)
 
 def play_game(player1, player2, state, alphazero: AlphaZero, first: bool):
 
@@ -42,7 +69,11 @@ def play_game(player1, player2, state, alphazero: AlphaZero, first: bool):
             print(state)
 
 
-def main(context: GameContext, first: bool):
+def main(context: GameContext, first: bool, mcts: bool = False):
+
+    if mcts:
+        play_mcts_vs_alphazero(context, first)
+        return
     alphazero = AlphaZero(context=context)
     state = alphazero.context.game.new_initial_state()
     play_game(player, ai, state, alphazero, first)