Why don't the developers of AlphaZero let it learn for days, or even months on end? Wouldn't its neural network become much stronger, at the cost of just occupying some computational resources for a while? It seems like this would make it indisputably the best engine in world, since right now there's still controversy on whether it's clearly better than Stockfish.

If something like this already is happening, please correct me.


Google's hardware is already very convincing. However, allowing a network running forever doesn't mean it'll play stronger chess. There is always limitation, in machine learning we say the network has reached "convergence".

Google developers must eventually stop the training, and evaluate it's performance. They would make a better model in the next iteration.

  • 1
    Do Alpha Zero developers feel that it reached convergence? – Akavall Feb 8 at 4:11
  • It's not "Google's" hardware and developers technically. They belong to a company called Deepmind. It has been bought by Google, but it's still a separate company. – Federico Poloni Feb 13 at 10:09

Check out the AlphaZero paper, figure 1.

enter image description here

As you can see from the leftmost figure, you can let AlphaZero train, but that doesn't mean it'll improve. In fact, AlphaZero peaked at a level slightly above Stockfish 8 (in other words, it'll likely lose to Stockfish 10).

I'm no expert, but I understand Leela is competitive with Stockfish 10 not because she has trained for longer, but because she's doing certain things that AlphaZero didn't do (such as "squeeze excitation" as in the latest Test 40).

  • Thanks for posting these figures, the visualization of AlphaZero's limitation is very insightful. – Inertial Ignorance Feb 9 at 0:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.