Komodo MCTS

[Ron Langeveld]

It has been a silent few months for Komodo to make progress on MCTS.
I was wondering whether or not Mark and Larry deem MCTS ready to have a go at the next TCEC Premier division instead of the standard Komodo?
The latter one is not likely to pose any real danger to Stockfish so why not have their new brainchild have run to see where it stands?

Ron

[jkiliani]

Komodo MC was recently disqualified from CCCC for apparent stability issues, and Chess.com has an actually vested interest in seeing it succeed. Even while the CCCC tests were still running, it didn't reach anywhere close to the strength of standard Komodo. From this I consider it rather unlikely that Komodo MC will replace Komodo in TCEC anytime soon. Entering Div 4 as a separate engine would make more sense to be honest, hopefully this will be allowed by TCEC.

[Branko Radovanovic]

My feelings too, and Komodo MCTS is my #1 wish for TCEC S14. In fact, I'd really like to see a non-NN MCTS engine entering. Apart from Komodo, there's only Scorpio, but it's not strong enough even for Div4.

[Ron Langeveld]

Stability issues due to MC developments? That's a bit troublesome! Since MC is integrated in the same exe I take it that subscribers will have to wait longer for the next installment of Komodo standard as well.

[lkaufman]

Current Komodo MCTS is much stronger than last release, maybe we have cut the gap from normal Komodo (on 1 or 4 threads) roughly in half since then. But we haven't yet found a way to make use of large numbers of cores, so having it play in TCEC or the chess.com event doesn't make much sense until we solve this problem. We have also had occasional crashes running on large numbers of threads. We could easily do a release now keeping a thread limit of something like 15 for MCTS, but we're hoping to solve this problem first. For 99% of our customers, this thread limit is no problem, so we might release without a solution, but it's a big problem for tournaments. It's not a big problem for our upcoming match with MVL, because we're using my fast 16 core machine rather than a slower one with 32 or more cores.