Discussion of chess software programming and technical issues.

Moderators: hgm, Dann Corbit, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
Posts: 13
Joined: Fri Mar 01, 2019 11:46 am
Full name: Marc-Philippe HUGET


Post by mphuget » Fri Dec 04, 2020 6:57 pm

Hello all,

In case you missed it, there is a new kid in town: Maiachess (
using a Deep Neural Network like AlphaZero, Leela, or others but does not rely on self-play.

The main interest in this one is to be trained at different levels (different ELO) then see what blunders
could do a 1100ELO comparing to ELO1900. Very informative in my opinion


Posts: 64
Joined: Sat Nov 14, 2020 11:49 am
Full name: Elias Nilsson

Re: Maiachess

Post by eligolf » Sat Dec 05, 2020 8:48 am

Yes I have seen this thank to Agadmator on YouTube. Looks like a very cool idea indeed.

I like the concept to be able to play against a human like opponent at different levels. Most of today's engines are way too strong to be able to play a fun game with. They are mainly used for analysing positions. This Maiachess looks like a fun opponent too. And the occasional blunders/mistakes makes it possible to recover even from losing positions, something that is very hard against a normal engine.

Post Reply