Skip to content
Chess reinforcement learning by AlphaGo Zero methods.
Jupyter Notebook Python
Branch: master
Clone or download
Zeta36 Merge pull request #80 from it-is-over9000/master
Made changes to to make self play work
Latest commit a04ec44 Apr 22, 2019
Type Name Latest commit message Commit time
Failed to load latest commit information.
binder adding gpu free environment and binder link Mar 5, 2018
data/model Updated best model after more supervised learning Jul 7, 2018
notebooks adding simple notebook, updating binder links Mar 5, 2018
src successfully changed a file to make the program work Apr 16, 2019
.gitignore - I upload the best model I was able to train until now. Dec 17, 2017
C0uci.bat works Dec 15, 2017
LICENSE.txt First commit Nov 19, 2017
Untitled.ipynb started training on aws, did not work Apr 15, 2019
model.png adding simple notebook, updating binder links Mar 5, 2018
requirements.txt - Lot of changes from @Akababa. Dec 24, 2017
test3.pgn started training on aws, did not work Apr 15, 2019

Binder Demo Notebook


Chess reinforcement learning by AlphaGo Zero methods.

This project is based on these main resources:

  1. DeepMind's Oct 19th publication: Mastering the Game of Go without Human Knowledge.
  2. The great Reversi development of the DeepMind ideas that @mokemokechicken did in his repo:
  3. DeepMind just released a new version of AlphaGo Zero (named now AlphaZero) where they master chess from scratch: In fact, in chess AlphaZero outperformed Stockfish after just 4 hours (300k steps) Wow!

See the wiki for more details.


I'm the creator of this repo. I (and some others collaborators did our best: but we found the self-play is too much costed for an only machine. Supervised learning worked fine but we never try the self-play by itself.

Anyway I want to mention we have moved to a new repo where lot of people is working in a distributed version of AZ for chess (MCTS in C++):

Project is almost done and everybody will be able to participate just by executing a pre-compiled windows (or Linux) application. A really great job and effort has been done is this project and I'm pretty sure we'll be able to simulate the DeepMind results in not too long time of distributed cooperation.

So, I ask everybody that wish to see a UCI engine running a neural network to beat Stockfish go into that repo and help with his machine power.


  • Python 3.6.3
  • tensorflow-gpu: 1.3.0
  • Keras: 2.0.8

New results (after a great number of modifications due to @Akababa)

Using supervised learning on about 10k games, I trained a model (7 residual blocks of 256 filters) to a guesstimate of 1200 elo with 1200 sims/move. One of the strengths of MCTS is it scales quite well with computing power.

Here you can see an example where I (black) played against the model in the repo (white):


Here you can see an example of a game where I (white, ~2000 elo) played against the model in this repo (black):


First "good" results

Using the new supervised learning step I created, I've been able to train a model to the point that seems to be learning the openings of chess. Also it seems the model starts to avoid losing naively pieces.

Here you can see an example of a game played for me against this model (AI plays black):


Here we have a game trained by @bame55 (AI plays white):


This model plays in this way after only 5 epoch iterations of the 'opt' worker, the 'eval' worker changed 4 times the best model (4 of 5). At this moment the loss of the 'opt' worker is 5.1 (and still seems to be converging very well).


Supervised Learning

I've done a supervised learning new pipeline step (to use those human games files "PGN" we can find in internet as play-data generator). This SL step was also used in the first and original version of AlphaGo and maybe chess is a some complex game that we have to pre-train first the policy model before starting the self-play process (i.e., maybe chess is too much complicated for a self training alone).

To use the new SL process is as simple as running in the beginning instead of the worker "self" the new worker "sl". Once the model converges enough with SL play-data we just stop the worker "sl" and start the worker "self" so the model will start improving now due to self-play data.

python src/chess_zero/ sl

If you want to use this new SL step you will have to download big PGN files (chess files) and paste them into the data/play_data folder (FICS is a good source of data). You can also use the SCID program to filter by headers like player ELO, game result and more.

To avoid overfitting, I recommend using data sets of at least 3000 games and running at most 3-4 epochs.

Reinforcement Learning

This AlphaGo Zero implementation consists of three workers: self, opt and eval.

  • self is Self-Play to generate training data by self-play using BestModel.
  • opt is Trainer to train model, and generate next-generation models.
  • eval is Evaluator to evaluate whether the next-generation model is better than BestModel. If better, replace BestModel.

Distributed Training

Now it's possible to train the model in a distributed way. The only thing needed is to use the new parameter:

  • --type distributed: use mini config for testing, (see src/chess_zero/configs/

So, in order to contribute to the distributed team you just need to run the three workers locally like this:

python src/chess_zero/ self --type distributed (or python src/chess_zero/ sl --type distributed)
python src/chess_zero/ opt --type distributed
python src/chess_zero/ eval --type distributed


  • uci launches the Universal Chess Interface, for use in a GUI.

To set up ChessZero with a GUI, point it to C0uci.bat (or rename to .sh). For example, this is screenshot of the random model using Arena's self-play feature: capture


  • data/model/model_best_*: BestModel.
  • data/model/next_generation/*: next-generation models.
  • data/play_data/play_*.json: generated training data.
  • logs/main.log: log file.

If you want to train the model from the beginning, delete the above directories.

How to use


install libraries

pip install -r requirements.txt

If you want to use GPU, follow these instructions to install with pip3.

Make sure Keras is using Tensorflow and you have Python 3.6.3+. Depending on your environment, you may have to run python3/pip3 instead of python/pip.

Basic Usage

For training model, execute Self-Play, Trainer and Evaluator.

Note: Make sure you are running the scripts from the top-level directory of this repo, i.e. python src/chess_zero/ opt, not python opt.


python src/chess_zero/ self

When executed, Self-Play will start using BestModel. If the BestModel does not exist, new random model will be created and become BestModel.


  • --new: create new BestModel
  • --type mini: use mini config for testing, (see src/chess_zero/configs/


python src/chess_zero/ opt

When executed, Training will start. A base model will be loaded from latest saved next-generation model. If not existed, BestModel is used. Trained model will be saved every epoch.


  • --type mini: use mini config for testing, (see src/chess_zero/configs/
  • --total-step: specify total step(mini-batch) numbers. The total step affects learning rate of training.


python src/chess_zero/ eval

When executed, Evaluation will start. It evaluates BestModel and the latest next-generation model by playing about 200 games. If next-generation model wins, it becomes BestModel.


  • --type mini: use mini config for testing, (see src/chess_zero/configs/

Tips and Memory

GPU Memory

Usually the lack of memory cause warnings, not error. If error happens, try to change vram_frac in src/configs/,

self.vram_frac = 1.0

Smaller batch_size will reduce memory usage of opt. Try to change TrainerConfig#batch_size in MiniConfig.

You can’t perform that action at this time.