TalkChess.com

I updated the code. Refactored a lot of it to try to make it easier to use in multiple different games and have some more standard search features and faster neural net calculation. GuiNN 2.04 against GuiCheckers 1.11 (an old pre-NN version) at .2s per-move 1169-31-1680 W-L-D (+145 elo) Which is pre...

Thanks for re-testing, at least it inched in the right direction this time. Hopefully will make more difference against stronger opponents.
2.4 is mostly just 2.3 with the endgame nets added. (I guess maybe had worse luck than 2.3 first test.)

Thanks for testing SlowChess 4CPU. When I re-started work on it I never expected to see it top 5 overall in the CCRL list.
(I expect position will start to drop now but it was a good run.)

I will probably try it out sometime this week in the checkers program and see if it seems worth it. It is more state to track, and the speed up is unlikely to be that big, but I expect could find a clear speedup for early/mid game. (To compute the first first layer values, currently the Checkers net...

This is really amazing! I've managed to both compile your engine and run your Python training code, it all works flawlessly. I have a GTX 1050 Ti with 4Gb RAM as GPU and that allows me to use a batch size of 1M (instead of your 15K). All the other parameter (learning schedule) I left untouched. Tra...

The result with Slow 2.4 slightly below 2.3 and 2.2 does come as a surprise even knowing the gains might not show up so much against weaker opponents. In SPCC and in the CEGT forum I see a 20+ elo gain for 2.4 vs 2.3, and versus the older 2.2 I got +42 elo standard, +51 FRC in my own testing. Not ha...

While cleaning up the GuiNN checkers code some more I managed to improve it by an extra 10 elo. So it definitely has some room to grow but maybe I am done now since checkers could be mostly covered by large databases instead, and not using them feels a bit artificial. International Draughts is an in...

Thanks for testing, tough second half for SlowChess. This is the first time I've seen someone use the own book in a match. I should start updating it again, I stopped around 1.8. I used to run 2min+1s matches with release candidates, and have it learn the first 13 moves of winning/drawing side(s). R...

Despite not working on chess, I have still been testing out neural nets. I started first with something very similar : Checkers. 15-20 years ago I had worked a bit on a checkers program. Using the same neural net code and training method I had used for chess endgames and detailed in this thread, I w...

Thanks for testing, +23 elo is slightly above my +21 self-play elo gain so I'm pretty happy with that result in a general rating list.
Probably/hopefully that means the endgame neural net learning scales better than some general gains do. (My testing was at 15s +.15)