Testing strategies for my engines playing strength

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

User avatar
lithander
Posts: 881
Joined: Sun Dec 27, 2020 2:40 am
Location: Bremen, Germany
Full name: Thomas Jahn

Re: Testing strategies for my engines playing strength

Post by lithander »

jdart wrote: Mon Feb 08, 2021 4:52 pm Btw. also you can get an engine account on one of the chess servers (chessclub.com or freechess.org for example). It can take some time/effort because you need to have a dedicated account for the engine that is registered as a computer account. But playing there is a good way to find bugs also.
Oh, that sounds fun! I will definitely look into that.
Guenther wrote: Mon Feb 08, 2021 8:19 pm Oops, I totally overlooked the <elo setting> for SF in your post, you need to look up the code [...]
I set UCI_LimitStrength option to 'false' and now SF has the expected 100% winrate! Thanks :)
Guenther wrote: Mon Feb 08, 2021 8:19 pm Well, actually I don't believe in testing against 'downscaled' programs, I would always try to find programs for testing who really play around the same strength. But that's just my opinion, others might disagree.
That's true. Luckily some online rankings also include lower strength engines so it won't be difficult to find opponents in my range. Actually, my hope is that with the next version MinimalChess will be stable and feature complete enough that contributors to rating lists such as CCRL can be compelled to add it. But first things first... back to working on time control issues. ;)
Minimal Chess (simple, open source, C#) - Youtube & Github
Leorik (competitive, in active development, C#) - Github & Lichess