Search found 212 matches

by frankp
Fri Sep 21, 2018 1:33 pm
Forum: Computer Chess Club: General Topics
Topic: 2080 Ti
Replies: 10
Views: 2536

Re: 2080 Ti

Someone posted nps numbers for fp16 and fp32 on the leela discord. Very impressive for the 2080Ti - about a titan v. fp32 not as good, but these are not used by leela - as I understand it. Working from memory, but I recall around 32 knps from the start position for fixed nodes (1M maybe). Vague reco...
by frankp
Tue Sep 18, 2018 10:36 am
Forum: Computer Chess Club: General Topics
Topic: Beautiful game, highlights strengths of NN+MCTS
Replies: 17
Views: 3094

Re: Beautiful game, highlights strengths of NN+MCTS

Does not change the beauty of the games, of course, which is not in doubt. But I asked the technical point about MCTS based on the paragraph below form the leela FAQ - which I guess is overseen by the leela devs. I guess the point is that it is not 'classic' MCTS as used by komodo (I believe). "This...
by frankp
Tue Sep 18, 2018 8:23 am
Forum: Computer Chess Club: General Topics
Topic: Beautiful game, highlights strengths of NN+MCTS
Replies: 17
Views: 3094

Re: Beautiful game, highlights strengths of NN+MCTS

Just on a technical point, does Lc0 actually do MCTS?
(The 4C version of Komodo does I believe. Not sure about Houdini.).
by frankp
Fri Sep 14, 2018 12:48 pm
Forum: Computer Chess Club: General Topics
Topic: Chess.com 2018 computer chess championship
Replies: 279
Views: 43656

Re: Chess.com 2018 computer chess championship

tomitank wrote:
Fri Sep 14, 2018 11:27 am
The Fire vs Lc0 and Komodo vs Lc0 etc.. games proved that artificial intelligence (NN in Chess) is not 100% trustworthy!
I was watching st louis 960 competition and certainly those chess-playing-monkeys are not be trusted either :)
by frankp
Thu Sep 13, 2018 8:45 am
Forum: Computer Chess Club: General Topics
Topic: leela is official(?) better than sf9
Replies: 81
Views: 10968

Re: leela is official(?) better than sf9

"The conclusion would be: AlphaZero team choose on purpose openings suitable to AlphaZero to boost significantly the performance, by more than 100 Elo points" Is a very strong accusation. I thought there was no opening book at all. Which meant, as with the other aspects of the match, SF8 was not pla...
by frankp
Fri Sep 07, 2018 9:25 pm
Forum: Computer Chess Club: General Topics
Topic: CCCC Lc0-Arasan : strange endgame
Replies: 7
Views: 1216

Re: CCCC Lc0-Arasan : strange endgame

lc0 is not using tb in 4C. The version playing is capable of only using WDL tables, but that makes the late endgame play super weird. The plan is to use both DTZ and WDL together once the code has been properly tested. In the next tournament, I guess, as changes to the engines are not allowed during...
by frankp
Tue Sep 04, 2018 2:36 pm
Forum: Computer Chess Club: General Topics
Topic: CCCC Rapid Rumble results simulator
Replies: 60
Views: 6957

Re: CCCC Rapid Rumble results simulator

Shredder has yet to play any of the top four, discounting itself of course.
So the real 4C table is probably still unstable - from an uncertainty viewpoint.
Ditto leela, komodo and houdini to a lesser extent.
by frankp
Mon Sep 03, 2018 1:11 pm
Forum: Computer Chess Club: General Topics
Topic: CCCC Rapid Rumble results simulator
Replies: 60
Views: 6957

Re: CCCC Rapid Rumble results simulator

that should help fill out the chess.com 'guess the final positions' competition :wink: . Out of interest, how did you score the two lc0 adjudicated games (for example, there were other crashes too), in terms of the chess performance to aid the prediction: +1 against ivanhoe and 0.5 gull? I guess pre...
by frankp
Mon Aug 20, 2018 9:11 am
Forum: Computer Chess Club: General Topics
Topic: AMD Threadripper 2990x seems out what is the gain vs 1950x ?
Replies: 38
Views: 5687

Re: AMD Threadripper 2990x seems out what is the gain vs 1950x ?

I thought I read somewhere that BMI is 'only' 32bit.
Certainly popcnt faster on Ryzen.
by frankp
Sun Aug 19, 2018 7:02 am
Forum: Computer Chess Club: General Topics
Topic: Strange Lc0 TCEC performance
Replies: 132
Views: 12484

Re: Strange Lc0 TCEC performance

I am still on the page that a self-trained neural network (probably not optimised for the task) given ‘only the rules of chess’ can in a few months hold its own against good AB engines, which are based on decades of research and development optimising search and the evaluation of terminal nodes. Th...