Search found 1631 matches

by yanquis1972
Mon Jan 28, 2019 4:43 pm
Forum: Computer Chess Club: General Topics
Topic: LC0 bad search and poor tactics
Replies: 45
Views: 6156

Re: LC0 bad search and poor tactics

t30 seems to have issues with basic QP opening positions, & overly optimistic pawn grabs at b2/g7...in a 20 game match at 10'+10s this was the only game lost https://lichess.org/study/CpmlY6g4/VtKMT2ow#0 absolutely smashed with no effort. what's more the slav was at one point a strength for t30, but...
by yanquis1972
Sun Jan 27, 2019 5:14 pm
Forum: Computer Chess Club: General Topics
Topic: LC0 bad search and poor tactics
Replies: 45
Views: 6156

Re: LC0 bad search and poor tactics

I think Lc0 is indeed very weak in tactics when compared to modern AB engines. I was looking at some of the games of a tourney I made against latest Dev SF and it lost many points by overlooking 1 move simple tactics. But even with this handicap it still trounced SF by around 90 ELO points. The way...
by yanquis1972
Sun Jan 27, 2019 3:21 am
Forum: Computer Chess Club: General Topics
Topic: Big boost in strength for LC0 by simply changing VGA card
Replies: 19
Views: 3457

Re: Big boost in strength for LC0 by simply changing VGA card

suspect there's been a significant gain in conjunction with the drop in self-elo, which seems to've recovered. a lot of complex endgames seem soft-solved; winning scores (~75%/~+2.50) where SF might give half/three quarters pawn advantage with fairly deep search. this 1min+1s bullet game on weak har...
by yanquis1972
Sat Jan 26, 2019 9:51 pm
Forum: Computer Chess Club: General Topics
Topic: LC0 bad search and poor tactics
Replies: 45
Views: 6156

Re: LC0 bad search and poor tactics

comparison test30 & test10 from move 7 (t30 seemed to fare no better w/ reductive FPU; i don't know what the recommended setting is) -- [D] rnbqk2r/ppp2ppp/4p3/8/1bBPn3/2N2N2/PP3PPP/R1BQ1RK1 b kq - 1 7 Lc0 id32906 [FPU: absolute]: 1/2 00:00.032 2 62 -48.45 7. ... Bxc3 8.bxc3 2/3 00:00.043 4 93 -50.2...
by yanquis1972
Sat Jan 26, 2019 5:52 pm
Forum: Computer Chess Club: General Topics
Topic: LC0 bad search and poor tactics
Replies: 45
Views: 6156

Re: LC0 bad search and poor tactics

yeah, what i found when looking at test10 & test40 in this position isn't that they avoid f6 (although it may be that t10 (11248) was able to see it was a mistake eventually; i don't recall) but that, unlike t30, avoid getting themselves into the position in the first place. which is another reason ...
by yanquis1972
Sat Jan 26, 2019 3:01 pm
Forum: Computer Chess Club: General Topics
Topic: LC0 bad search and poor tactics
Replies: 45
Views: 6156

Re: LC0 bad search and poor tactics

So spend $1200 dollars, create his own networks, and play something other than chess.
by yanquis1972
Sat Jan 26, 2019 2:57 pm
Forum: Computer Chess Club: General Topics
Topic: Novelty in the Sicilian Najdorf and interesting plan of Lc0 AF228
Replies: 4
Views: 1008

Re: Novelty in the Sicilian Najdorf and interesting plan of Lc0 AF228

I recall seeing some KS attacks with test30 some time ago; haven’t been able to observe many games so I don’t know if it’s still doing it or playing in a more traditional manner. I think it will always try to shape the game to favored patterns so eg a net that likes or is trained on the KID (somethi...
by yanquis1972
Sat Jan 26, 2019 1:26 am
Forum: Computer Chess Club: General Topics
Topic: LC0 bad search and poor tactics
Replies: 45
Views: 6156

Re: LC0 bad search and poor tactics

Vs SF10 (Bxc4 forced)— [Event "Computer chess game"] [Site "PC"] [Date "2019.01.25"] [Round "?"] [White "Stockfish 10"] [Black "Lc0 T30"] [Result "*"] [BlackElo "2980"] [ECO "D24"] [Opening "QGA"] [Time "18:58:39"] [Variation "4.Nc3 e6 5.e4"] [WhiteElo "3000"] [TimeControl "300"] [Termination "unter...
by yanquis1972
Sat Jan 26, 2019 1:13 am
Forum: Computer Chess Club: General Topics
Topic: LC0 bad search and poor tactics
Replies: 45
Views: 6156

Re: LC0 bad search and poor tactics

The resonant point is that, as dann points out, this is a present issue with test30. The nets are not only blundering 7 moves into a a game, but from a basic position that arises from one of deepminds test set — or simply 1. Nf3, as below. Example with 32891, me forcing moves after a few seconds (I ...