What ELO rating was the engine able to achieve so far

Zeta36 / chess-alpha-zero

Chess reinforcement learning by AlphaGo Zero methods.

MIT License

2.13k stars 481 forks source link

What ELO rating was the engine able to achieve so far #66

Open salvador-dali opened 6 years ago

salvador-dali commented 6 years ago

In the description of the project I was not able to find the current ELO rating of the engine achieved so far (also on what hardware and under what conditions was it achieved).

Have I missed this information somewhere?

brianprichardson commented 6 years ago

I don't think things got far enough with this project to post an Elo number. Again, Leela Chess Zero is further along, but even there they are just starting to get some comparable Elo's with other engines after making sure the NN beats random play and as it improves can keep beating itself.

Akababa commented 6 years ago

We haven't measured it, but if I had to guess I wouldn't say it's more than 1200. It takes an absurd amount of resources to make this work well and this is mostly an instructional/exploratory project. As @brianprichardson points out the C++ project is more serious but I'm not sure if they have decent results yet.