cleanup and results

.github/workflows/python-app.yml

@@ -1,39 +1,39 @@
 # This workflow will install Python dependencies, run tests and lint with a single version of Python
 # For more information see: https://docs.github.com/en/actions/automating-builds-and-tests/building-and-testing-python
 
-name: Python application
+# name: Python application
 
-on:
-  push:
-    branches: [ "main" ]
-  pull_request:
-    branches: [ "main" ]
+# on:
+#   push:
+#     branches: [ "main" ]
+#   pull_request:
+#     branches: [ "main" ]
 
-permissions:
-  contents: read
+# permissions:
+#   contents: read
 
-jobs:
-  build:
+# jobs:
+#   build:
 
-    runs-on: ubuntu-latest
+#     runs-on: ubuntu-latest
 
-    steps:
-    - uses: actions/checkout@v3
-    - name: Set up Python 3.10
-      uses: actions/setup-python@v3
-      with:
-        python-version: "3.10"
-    - name: Install dependencies
-      run: |
-        python -m pip install --upgrade pip
-        pip install pytest
-        if [ -f requirements.txt ]; then pip install -r requirements.txt; fi
-    - name: Evaluate
-      if: contains(github.event.head_commit.message, 'EVAL')
-      run: |
-        python -m ci_quixo
-    - name: Create Pull Request
-      uses: peter-evans/create-pull-request@v5
-      with:
-        title: Generate nuove foto di eval
-        commit-message: updated results picture
+#     steps:
+#     - uses: actions/checkout@v3
+#     - name: Set up Python 3.10
+#       uses: actions/setup-python@v3
+#       with:
+#         python-version: "3.10"
+#     - name: Install dependencies
+#       run: |
+#         python -m pip install --upgrade pip
+#         pip install pytest
+#         if [ -f requirements.txt ]; then pip install -r requirements.txt; fi
+#     - name: Evaluate
+#       if: contains(github.event.head_commit.message, 'EVAL')
+#       run: |
+#         python -m ci_quixo
+#     - name: Create Pull Request
+#       uses: peter-evans/create-pull-request@v5
+#       with:
+#         title: Generate nuove foto di eval
+#         commit-message: updated results picture

README.md

@@ -18,11 +18,40 @@ The game Quixo is a Tic-Tac-Toe variant, played on a five-by-five board of cubes
 ## Navigate through the code
 
 - `custom_game.py`: Wrapper around Game class, with some utility methods and symmetry (canonical representation) handling
-- `minmax.py` and `mcts.py`: Player's files
+- `minmax.py`: Minmax implementation
+  - Features:
+    - Depth-Limited
+    - Alpha-Beta Pruning
+    - Hash-Tables
+    - Different Pruning Levels:
+      - Pruning lvl 0 - No Pruning, at each node consider all possible moves
+      - Pruning lvl 1 - At each node consider only the moves that land on different boards 
+      - Pruning lvl 2 - At each node consider only the moves that land on different boards and that are not symmetric to each other
+      - Pruning lvl 3 - At each node consider only the moves that land on different boards and that are not symmetric to each other and that are not have been already visited on a higher level (not really that useful since we limit to a depth of 3, so there really is not a chance for a cycle to happen)  
+- `mcts.py`: MonteCarlo Tree Search implementation
+  - Features:
+    - Random
+    - w/ Heuristic
 - `__main__.py`: containis the code to perform the evaluation
 
+## Results obtained
+
+### Vs Random
+
+|                    Agent                          | Games | Win Rate (%) | Average Time per Game (s) | Average Time per Move (s) | Total Number of Moves |
+|:-------------------------------------------------:|:-----:|:------------:|:-------------------------:|:-------------------------:|:---------------------:|
+|   MinMax(AB,D2,P1) - αβ + Depth 2 + Pruning lvl 1 |  100  |      100     |            0.61           |            0.07           |          831          |
+|   MinMax(AB,D3,P2) - αβ + Depth 3 + Pruning lvl 2 |  100  |      100     |           11.97           |            1.47           |          816          |
+|      MCTS(R500) - 500 Games with Random Moves     |  100  |      84      |            8.91           |            0.90           |          985          |
+|        MCTS(H500) - 500 Games with Heuristic      |  100  |      79      |           52.45           |            4.42           |          1186         |
+
+![Win Rates](./ci_quixo/results/players_wr.png)
+
+![Time Comparison](./ci_quixo/results/time_comparison.png)
+
 ## Possible Improvements
 
 - [ ] Minmax w/ RankCut
 - [ ] Minmax w/ Singular Moves (should be easy and fast-enough using Hash Tables)
-- [ ] Minmax w/ Parallelization
+- [ ] Parallelization both on Minmax and MCTS
+- [ ] Lookup Tables for starting positions

ci_quixo/__main__.py

@@ -6,14 +6,15 @@
     from ._players import *
     from .main import Player
     from .helper import evaluate, Result, gen_plots
-from pprint import pprint
 import dill, multiprocessing
 dill.Pickler.dumps, dill.Pickler.loads = dill.dumps, dill.loads
 multiprocessing.reduction.ForkingPickler = dill.Pickler
 multiprocessing.reduction.dump = dill.dump
 
-GAMES = 10
+GAMES = 100
 HIDE_PBAR = True
+PARALLELIZE = True
+DISPLAY_RES = True
 
 minmax2 = MinMaxPlayer()
 minmax3 = MinMaxPlayer(3, pruning=2)
@@ -22,32 +23,34 @@
 
 
 def test_minmax_vs_random() -> Result:
-    return evaluate(minmax2, RandomPlayer(), games=GAMES, display=True, hide_pbar=HIDE_PBAR)
+    return evaluate(minmax2, RandomPlayer(), games=GAMES, display=DISPLAY_RES, hide_pbar=HIDE_PBAR)
 
 def test_minmax3_vs_random() -> Result:
-    return evaluate(minmax3, RandomPlayer(), games=GAMES, display=True, hide_pbar=HIDE_PBAR)
+    return evaluate(minmax3, RandomPlayer(), games=GAMES, display=DISPLAY_RES, hide_pbar=HIDE_PBAR)
 
 def test_mcts_r_vs_random() -> Result:
-    return evaluate(mcts_r, RandomPlayer(), games=GAMES, display=True, hide_pbar=HIDE_PBAR)
+    return evaluate(mcts_r, RandomPlayer(), games=GAMES, display=DISPLAY_RES, hide_pbar=HIDE_PBAR)
 
 
 def test_mcts_h_vs_random() -> Result:
-    return evaluate(mcts_h, RandomPlayer(), games=GAMES, display=True, hide_pbar=HIDE_PBAR)
+    return evaluate(mcts_h, RandomPlayer(), games=GAMES, display=DISPLAY_RES, hide_pbar=HIDE_PBAR)
 
 
 def test_minmax_vs_mcts_random() -> Result:
-    return evaluate(minmax2, mcts_r, games=GAMES, display=True, hide_pbar=HIDE_PBAR)
+    return evaluate(minmax2, mcts_r, games=GAMES, display=DISPLAY_RES, hide_pbar=HIDE_PBAR)
 
 def test_minmax_vs_mcts_heuristic() -> Result:
-    return evaluate(minmax2, mcts_h, games=GAMES, display=True, hide_pbar=HIDE_PBAR)
+    return evaluate(minmax2, mcts_h, games=GAMES, display=DISPLAY_RES, hide_pbar=HIDE_PBAR)
 
 def call(it: callable) -> Result:
     return it()
 
 if __name__ == "__main__":
-    evals = [test_minmax_vs_random, test_minmax3_vs_random, test_mcts_r_vs_random, test_mcts_h_vs_random, test_minmax_vs_mcts_random, test_minmax_vs_mcts_heuristic]
-
-    with multiprocessing.Pool() as p:
-        RESULTS = p.map(call, evals[:-2])
-
+    from tqdm import tqdm
+    evals = [test_minmax_vs_random, test_minmax3_vs_random, test_mcts_r_vs_random, test_mcts_h_vs_random, test_minmax_vs_mcts_random, test_minmax_vs_mcts_heuristic][:]
+    if PARALLELIZE:
+        with multiprocessing.Pool() as p:
+            RESULTS = list(tqdm(p.imap(call, evals), total=len(evals), unit="test", desc=f"Testing with {GAMES} games each"))
+    else:
+        RES = [it() for it in tqdm(evals, desc="Evaluating", unit="eval")]        
     gen_plots(RESULTS)

ci_quixo/helper.py

@@ -10,7 +10,7 @@
     if TYPE_CHECKING:
         from .custom_game import CompleteMove
 
-import time
+import time, dill
 import numpy as np
 from tqdm.auto import trange
 from os import path
@@ -88,6 +88,8 @@ def evaluate(p1: "Player", p2: "Player" = None, games: int = 10, display: bool =
 
 
 def gen_plots(results: list[Result]) -> None:
+    with open(path.join(PLOT_FOLDER, "results.bkp.pkl"), "wb+") as f:
+        dill.dump(results, f)
     plot_time_comparison(results)
     plot_wr_vs_random(results)
 
@@ -107,8 +109,8 @@ def plot_wr_vs_random(results: list[Result]) -> None:
         "Overall": wr
     }
 
-    x = np.arange(len(player_names)) * 1.6
-    width = 0.30  # the width of the bars
+    x = np.arange(len(player_names)*2, step=2)
+    width = 0.50  # the width of the bars
 
 
     fig, ax = plt.subplots(layout='constrained')
@@ -122,7 +124,7 @@ def plot_wr_vs_random(results: list[Result]) -> None:
     # Add some text for labels, title and custom x-axis tick labels, etc.
     ax.set_ylabel('Win Rate (%)')
     ax.set_title('Agents Win Rate vs Random')
-    ax.set_xticks(x + width, player_names)
+    ax.set_xticks(x + width, player_names, rotation=-20)
     ax.legend(loc='upper left', ncols=3)
     ax.set_ylim(0, 120)
 
@@ -135,7 +137,7 @@ def plot_time_comparison(results: list[Result]) -> None:
     ys = [r.avg_time for r in results]
     xs = [r.p1.short_name for r in results]
     plt_title = "Average Seconds per Game (vs Random)"
-    plt.plot(xs, ys, 'o')
+    plt.plot(xs, ys, marker='o', linestyle='--')
     plt.title(plt_title)
     plt.ylabel("seconds")
     plt.yticks(ys)