Search found 216 matches

by mmt
Thu Apr 23, 2020 6:18 am
Forum: Computer Chess Club: General Topics
Topic: Automatically created test set of very difficult moves
Replies: 2
Views: 466

Re: Automatically created test set of very difficult moves

That does sound like a good way to do it.
by mmt
Wed Apr 22, 2020 12:09 am
Forum: Computer Chess Club: General Topics
Topic: Automatically created test set of very difficult moves
Replies: 2
Views: 466

Automatically created test set of very difficult moves

Is there something like this out there? It could be made by looking at positions where both SF and LC0 with 7-piece tablebases take more than 30 minutes on high-end hardware to switch their best move to a single different move and then also evaluate other all other moves as being at least 2.0 worse....
by mmt
Tue Apr 21, 2020 12:50 pm
Forum: Computer Chess Club: General Topics
Topic: Amazing tactic played by a middle school boy!
Replies: 6
Views: 818

Re: Amazing tactic played by a middle school boy!

I think Qe4 and Bb3 are also winning, so it's not great as a test position. Still, it's a nice move.
by mmt
Tue Apr 21, 2020 12:31 pm
Forum: Computer Chess Club: General Topics
Topic: Test position: Kasparov-Pribyl (1980) - very tough
Replies: 12
Views: 2484

Re: Test position: Kasparov-Pribyl (1980) - very tough

LC0 with 384x30-t60-3180 finds it but I haven't recorded in how much time (it's less than 4 hours).
by mmt
Wed Apr 15, 2020 3:50 pm
Forum: Computer Chess Club: Tournaments and Matches
Topic: For how long will Stockfish be competitive versus the best NN ?
Replies: 78
Views: 11515

Re: For how long will Stockfish be competitive versus the best NN ?

SF will be competitive until the LC0 team focuses on the tactics and the endgame. If SF can reach EGTBs because it looks deeper and LC0 cannot, it's a lost cause in the cases when the endgame is not straightforward. Learning on games with a larger variety of openings will also help.
by mmt
Tue Apr 14, 2020 1:11 am
Forum: Computer Chess Club: General Topics
Topic: Test position bm Bxh2+
Replies: 23
Views: 1768

Re: 2nd test position (Black to play) bm Nfd5!!

MikeGL wrote:
Mon Apr 13, 2020 5:33 pm
LC0 sees the right move instantly but it gets to only +1.25 in one minute. +1.54 after 10 minutes.
by mmt
Fri Apr 03, 2020 5:43 am
Forum: Computer Chess Club: General Topics
Topic: Fairly difficult EG puzzle
Replies: 21
Views: 1813

Re: Fairly difficult EG puzzle

LC0 0.24.1 with 384x30-t60-3180 gets Nh7+ with a positive score in 9 seconds on my system (Titan V + 2080Ti). With 63053 it finds Nh7+ with a positive score in 6 seconds.
by mmt
Fri Apr 03, 2020 5:15 am
Forum: Computer Chess Club: General Topics
Topic: Banksia GUI released
Replies: 842
Views: 95831

Re: Banksia GUI released

Actually BSG did not hard code parameters of the draw adjudication thus users can set them manually. However, users need to modify its JSON file, called banksiatour.json. This file is in BSG's configuration folder (in which BSG stored all internal data). Will you add this to the GUI? I think those ...
by mmt
Fri Mar 27, 2020 11:11 am
Forum: Computer Chess Club: General Topics
Topic: 100 Elo drop for LC0?
Replies: 3
Views: 756

Re: 100 Elo drop for LC0?

Maybe not the greatest GPU combined with not very long time controls but seems weaker than expected even so.
by mmt
Mon Mar 23, 2020 9:33 pm
Forum: Computer Chess Club: General Topics
Topic: Issue with compilation of leela ver.0.24.x
Replies: 25
Views: 2180

Re: Issue with compilation of leela ver.0.24.x

When I was compiling 0.23.2, in the VS solution there was a directory subprojects and protobuf-lite is there. You have to build it.