Search found 3746 matches

by jdart
Sun Jan 12, 2020 2:28 am
Forum: Computer Chess Club: General Topics
Topic: interesting position
Replies: 6
Views: 421

Re: interesting position

Alayan wrote:
Sun Jan 12, 2020 2:21 am
jdart wrote:
Sun Jan 12, 2020 2:11 am
Black's King is in some distress, but on the other hand, White has a couple of passers.
You mean white's king ?
Yes
by jdart
Sun Jan 12, 2020 2:11 am
Forum: Computer Chess Club: General Topics
Topic: interesting position
Replies: 6
Views: 421

interesting position

From a recent online game: 5rk1/5pp1/pQ3n1p/2P5/2pP4/P1Nb2P1/1B3PKP/4r3 w - - Black's King is in some distress, but on the other hand, White has a couple of passers. It is even, but White has to play the somewhat unintuitive g4! here. Stockfish and Houdini are very fast on this but other engines may...
by jdart
Sat Jan 11, 2020 6:59 pm
Forum: Computer Chess Club: General Topics
Topic: Ryzen 3700x build
Replies: 10
Views: 613

Re: Ryzen 3700x build

Modern Times wrote:
Sat Jan 11, 2020 6:52 pm
64GB of RAM seems a lot, money could have been saved with much less ?
You could get by with less. My engine allocates hash in terms of powers of two, so 64G physical RAM allows at least 32G hash. You could use most of that hash in a long time-control game.

--Jon
by jdart
Sat Jan 11, 2020 6:34 pm
Forum: Computer Chess Club: General Topics
Topic: Ryzen 3700x build
Replies: 10
Views: 613

Ryzen 3700x build

I really want the 32-core or even 64-core Threadripper, but the 32-core is still scarce and the 64-core not out yet. Meanwhile I have built a Ryzen 3700x system. This will be mostly for running Arasan on ICC: MSI B450 Tomahawk Max motherboard CORSAIR Vengeance LPX 64GB (4 x 16GB) 288-Pin DDR4 SDRAM ...
by jdart
Fri Jan 10, 2020 12:42 pm
Forum: Computer Chess Club: General Topics
Topic: Skippers two month record
Replies: 12
Views: 1206

Re: Skippers two month record

Some thoughts: IMO most of the difference between very low-rated and higher-rated engines are due to bugs. Bugs can include common programming errors like array overrun, and also logic errors. It is really essential to have code that you know does what you intended it to do. The cure for this is tes...
by jdart
Sun Jan 05, 2020 7:13 pm
Forum: Computer Chess Club: Tournaments and Matches
Topic: FGRL 60 min + 15 sec Rating list - Schooner 2.2, Lc0 v0.23.1 LD2, Arasan 22.0
Replies: 3
Views: 365

Re: FGRL 60 min + 15 sec Rating list - Schooner 2.2, Lc0 v0.23.1 LD2, Arasan 22.0

I always find your matches interesting, especially the LTC ones. It looks to me though that you have some pretty unbalanced openings. For example, in this game: [Event "?"] [Site "?"] [Date "2019.10.26"] [Round "1"] [White "Stockfish 10"] [Black "Komodo 13.2"] [Result "1-0"] [ECO "E20"] [PlyCount "1...
by jdart
Sat Jan 04, 2020 10:07 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Perceptron for reduction / extension / other things
Replies: 3
Views: 479

Re: Perceptron for reduction / extension / other things

I think there is nothing wrong with it in theory. You can also use a black-box optimizer. CLOP (https://www.remi-coulom.fr/CLOP/) is one framework that can be used for this for example. I believe Stockfish and other programs have used SPSA (referenced in the CLOP paper), and I have tried a couple of...
by jdart
Tue Dec 31, 2019 4:32 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: RMO - Randomized Move Order - yet another Lazy SMP derivate
Replies: 113
Views: 6087

Re: RMO - Randomized Move Order - yet another Lazy SMP derivate

Most of modern alpha-beta search logic is just heuristics that work enough of the time to give an ELO gain. There are many positions where they will fail. I remember when Toga for example was one of the early engines making heavy use of late move pruning. That is just throwing away big chunks of the...
by jdart
Mon Dec 30, 2019 2:43 am
Forum: Computer Chess Club: General Topics
Topic: New Cerebellum Library
Replies: 52
Views: 5965

Re: New Cerebellum Library

FYI there were a couple of previous discussions on book move selection logic, see: http://talkchess.com/forum3/viewtopic.php?f=7&t=67744 and http://talkchess.com/forum3/viewtopic.php?f=2&t=71297 This is actually something that has some mathematical foundation. The optimal strategy is neither to sele...
by jdart
Sun Dec 29, 2019 2:25 am
Forum: Computer Chess Club: General Topics
Topic: New Cerebellum Library
Replies: 52
Views: 5965

Re: New Cerebellum Library

There are many cases where the best continuation is not clear. Just for one example, this position from the Petroff: rnbqkb1r/ppp2ppp/8/3p4/3Pn3/3B1N2/PPP2PPP/RNBQK2R b KQkq - 0 6 Black can play .. Nc6, .. Be7, .. Be6, or .. Bf5. Strong players, including correspondence players, still use all of the...