Search found 1631 matches

by yanquis1972
Sun Feb 03, 2019 4:48 pm
Forum: Computer Chess Club: General Topics
Topic: Lc0 with latest test30 nets is vastly superior positionally
Replies: 39
Views: 8258

Re: Lc0 with latest test30 nets is vastly superior positionally

i was wondering about scaling based on its CCC performance; hopefully that's the case. for whatever reason it's been my understanding that both t10 & t30 suffered in tactical tests over time (after first LR drop), so if that reverses it'd be huge as well. fwiw i got 153/200 with my settings; logs at...
by yanquis1972
Sun Feb 03, 2019 1:51 am
Forum: Computer Chess Club: General Topics
Topic: Lc0 with latest test30 nets is vastly superior positionally
Replies: 39
Views: 8258

Re: Lc0 with latest test30 nets is vastly superior positionally

Kai, have you tested t40 yet? I used drastically different parameters (60s/position, abandon after 10 or more plies+1 if best move) and stopped after 95 positions, but I wouldn’t be surprised if it’s already well ahead of AB engines. I can only guess a lot of essential knowledge from test35 was reta...
by yanquis1972
Sun Feb 03, 2019 1:42 am
Forum: Computer Chess Club: General Topics
Topic: Interesting game from CCC 5: Fire - Lc0
Replies: 8
Views: 1326

Re: Interesting game from CCC 5: Fire - Lc0

T40s endgame eval in general already seems phenomenal. Does anyone know of a good endgame test suite? I’m curious how it stacks up to T30 minus TBs.
by yanquis1972
Thu Jan 31, 2019 4:16 pm
Forum: Computer Chess Club: General Topics
Topic: Lc0 ... the GOOD and the BAD
Replies: 54
Views: 10621

Re: Lc0 ... the GOOD and the BAD

it depends on the net, probably -- i'm guessing m ansari was using t30 nets. nevertheless there's something to what you say. more current ones seem extremely likely to have played Qd5+ but interestingly the search actually hurts here; Qd5+ is immediately proposed as completely winning before eval dr...
by yanquis1972
Wed Jan 30, 2019 10:03 pm
Forum: Computer Chess Club: Tournaments and Matches
Topic: Super bullet chess: LC0 network 40445 (4000 selfplay elo) vs Stockfish 8
Replies: 4
Views: 1047

Re: Super bullet chess: LC0 network 40445 (4000 selfplay elo) vs Stockfish 8

yes, but my point is there's no point of comparison. it's +26 elo in that sample, with that particular configuration. what's test10? test30? it's clearly an excellent result for a young net but to say test40 is already ahead of SF8 (in comparison to any previous benchmark) is simply not true.
by yanquis1972
Tue Jan 29, 2019 4:15 pm
Forum: Computer Chess Club: General Topics
Topic: LC0 bad search and poor tactics
Replies: 45
Views: 6138

Re: LC0 bad search and poor tactics

another example, probably lost before move 8 -- [Event "Lc0 20 start"] [Site "PC"] [Date "2019.01.29"] [Round "4"] [White "SFDEV"] [Black "Lc0 T30 id32952"] [Result "*"] [BlackElo "3000"] [ECO "D10"] [Opening "Slav"] [Time "09:53:05"] [Variation "Exchange, 4.Bf4"] [WhiteElo "3000"] [TimeControl "600...
by yanquis1972
Tue Jan 29, 2019 2:46 pm
Forum: Computer Chess Club: General Topics
Topic: Did Leela miss a win in the KGA vs SF here?
Replies: 3
Views: 772

Did Leela miss a win in the KGA vs SF here?

[Event "SfDev vs lc0 id32952+"] [Site "PC"] [Date "2019.01.29"] [Round "1"] [White "Lc0 T30 id32952"] [Black "SFDEV"] [Result "1/2-1/2"] [BlackElo "3000"] [ECO "C33"] [Opening "KGA"] [Time "07:55:16"] [Variation "Bishop's Gambit"] [WhiteElo "3000"] [TimeControl "600+10"] [Termination "normal"] [Ply...
by yanquis1972
Mon Jan 28, 2019 8:21 pm
Forum: Computer Chess Club: General Topics
Topic: SF10 blunder??
Replies: 6
Views: 1051

Re: SF10 blunder??

Leelas eval is in win percentage, so it actually dropped after 42. Rc2 (SF also shows a slight plus for some reason) and shot up to 90% after 43 Rc1. I trust your judgment but Rc2 prepares Rf3 defending the pawn; I assume there are some tactical tricks that both engines missed?
by yanquis1972
Mon Jan 28, 2019 7:10 pm
Forum: Computer Chess Club: General Topics
Topic: SF10 blunder??
Replies: 6
Views: 1051

Re: SF10 blunder??

Arena with 5 man TBs, 4GB hash for SF (I suspect one of these two is the issue if it’s not replicable), otherwise should be default settings.

Used the code tag since I wasn’t sure if the pgn tag kept all the info.
by yanquis1972
Mon Jan 28, 2019 6:33 pm
Forum: Computer Chess Club: General Topics
Topic: SF10 blunder??
Replies: 6
Views: 1051

SF10 blunder??

anyone able to replicate this/explain the logic? SF just gives up a pawn in what both engines had evaluated as a drawish/drawn position with 43. Rc1... my CPU is currently occupied w/ the ongoing match so i'm curious if SF discovered the position was in fact completely lost. [d]8/1p1r1p2/1R2b3/P2p1k...