Gabor Szots wrote: ↑Mon May 10, 2021 12:00 pm
I guess it was not entirely a guess.
True. But I tested MMC only against 5 other engines of which only WukongJS was also used by CCRL. And I run all my games at 5s + 0.5s increment time controls. I often run so many games in parallel that my CPU sits at 100% load. And I didn't anchor any of the other engines and didn't let them play each other.
All that combined let's me test well whether a new version performs better than it's predecessor but didn't inspire much confidence in my ability to judge it's "general" playing strength under CCRL conditions.
Minimal Chess (simple, open source, C#) - Youtube & Github Leorik (competitive, in active development, C#) - Github & Lichess
...seems like the guesstimate of 1900 ELO was somewhat accurate!
I got something like 1875-1890 in my own testing, depending on which engines I'd use as anchors. 1887 is right in the middle of the error margin, so I think it's a great result. Congratulations.
Yesterday I took another step towards finishing the catproofing of the apartment, and the new Makefile is almost done, so I'll be resuming development of the engine in the near future. Why must I be catching up to the likes of you? (*)
(*) (Jinpachi Mishima in Tekken Tag 2 says: "Why must I be paired with the likes of you?" if he has to fight alongside a character he doesn't like. It's the only game at the moment I sometimes play.)