Search found 57 matches

by MonteCarlo
Sun Dec 16, 2018 5:52 pm
Forum: Computer Chess Club: General Topics
Topic: Alphazero news
Replies: 358
Views: 19495

Re: Alphazero news

As a side note, how confident are we that this wide variance with opening type of results for A0/LC0 is due to their learning the game in some very specific way and not due to, say, MCTS' being substantially weaker in certain types of positions? One maybe interesting thing to check would be to see i...
by MonteCarlo
Wed Dec 05, 2018 5:06 pm
Forum: Computer Chess Club: Tournaments and Matches
Topic: LC0 network 31740 vs Stockfish 10
Replies: 12
Views: 1302

Re: LC0 network 31740 vs Stockfish 10

Except your joke doesn't work, because to even run an OS and chess engines you need a CPU. A GPU alone can't even boot the computer. I can also beat Leela and Stockfish at Tetris and Pacman easily. I'll repeat what I already said to Milos and you quoted: Not my joke, and even if it were, if all jok...
by MonteCarlo
Tue Dec 04, 2018 5:46 pm
Forum: Computer Chess Club: Tournaments and Matches
Topic: LC0 network 31740 vs Stockfish 10
Replies: 12
Views: 1302

Re: LC0 network 31740 vs Stockfish 10

Your reply as well as the upper attempt of "joke" shows how ppl are actually clueless and lack basic logic (and the wonderful irony in your nickname that is actually name of algorithm that exclusively requires CPU to run). Neither SF nor Lc0 nor any other engine can actually run only on graphics ca...
by MonteCarlo
Tue Dec 04, 2018 3:55 pm
Forum: Computer Chess Club: Tournaments and Matches
Topic: LC0 network 31740 vs Stockfish 10
Replies: 12
Views: 1302

Re: LC0 network 31740 vs Stockfish 10

Lc0 Network 31777 Vs Stockfish 10 Both engines using GTX 965 Graphics Card with over 1000 CUDA Cores. 50 openings limited to 6 moves. TC: 1min+1s Result: Lc0 100 points Stockfish 10 0 points +100 =0 -0 Winning percentage = 100% Stockfish 10 is 1000 Elo weaker than Lc0. More LCO fake news ? Actually...
by MonteCarlo
Wed Nov 07, 2018 12:37 am
Forum: Computer Chess Club: General Topics
Topic: Komodo 12.2 Announcement
Replies: 54
Views: 5732

Re: Komodo 12.2 Announcement

Just another example along these lines, but more extreme. Maybe this will help shed some light: setoption name use mcts value true position startpos moves g2g4 e7e5 f2f4 go infinite info depth 8 seldepth 1 time 2001 score cp 2858469 nodes 54 nps 26 pv d8h4 info depth 10 seldepth 1 time 4003 score cp...
by MonteCarlo
Thu May 03, 2018 2:21 am
Forum: Computer Chess Club: General Topics
Topic: LCZero Accomplishments and Goals Thus Far
Replies: 177
Views: 20759

Re: LCZero Accomplishments and Goals Thus Far

Well, the search algorithm plays the move with the most visits. The initial visit is made to the move with the highest probability from the policy head. With a 1 visit "search", the search will evaluate the position after the move with the highest probability from the policy head, but no matter wha...
by MonteCarlo
Thu May 03, 2018 12:44 am
Forum: Computer Chess Club: General Topics
Topic: LCZero Accomplishments and Goals Thus Far
Replies: 177
Views: 20759

Re: LCZero Accomplishments and Goals Thus Far

Well, the search algorithm plays the move with the most visits. The initial visit is made to the move with the highest probability from the policy head. With a 1 visit "search", the search will evaluate the position after the move with the highest probability from the policy head, but no matter what...
by MonteCarlo
Wed May 02, 2018 11:38 pm
Forum: Computer Chess Club: General Topics
Topic: LCZero Accomplishments and Goals Thus Far
Replies: 177
Views: 20759

Re: LCZero Accomplishments and Goals Thus Far

For me it's exciting but it's not that incredible as I was following Leela Zero in go and something similar happened there. I like playing against pure policy networks (1 playout, no search) , it's a lot of fun in both chess and go. In go pure network is already around 3-4 dan level which would cor...
by MonteCarlo
Tue May 01, 2018 8:30 pm
Forum: Computer Chess Club: General Topics
Topic: Elo/source-code meter
Replies: 14
Views: 1421

Re: Elo/source-code meter

Should probably slightly change the original calculation. It's taking the expected score of Engine A vs Engine B, then doing 1/expected score to show much better Engine B is. When Engine A and Engine B have the same rating, that would output that each engine is twice as good as the other :) If you m...
by MonteCarlo
Tue May 01, 2018 8:21 pm
Forum: Computer Chess Club: General Topics
Topic: Leela on more then one GPU??
Replies: 1
Views: 358

Re: Leela on more then one GPU??

Yes, Leela supports multiple GPUs. You just pass multiple --gpu <OpenCL ID> parameters to the engine. I've tested that this works on 8 K80s (although I didn't have enough CPU to fully utilize all those GPUs, at 1 thread per GPU NPS scaled linearly; I just couldn't drive GPU utilization much past 50%...