Search found 398 matches

by jhellis3
Wed Nov 06, 2019 10:48 pm
Forum: Computer Chess Club: General Topics
Topic: Stockfish with and without contempt
Replies: 36
Views: 2003

Re: Stockfish with and without contempt

If you create a git branch with the change, you can just pull master into it and it will auto merge as long as no changes have been made to the contempt code upstream.
but I need the correct score.
Well, the chances you aren't getting that are about the same regardless.
by jhellis3
Wed Nov 06, 2019 10:28 pm
Forum: Computer Chess Club: General Topics
Topic: Stockfish with and without contempt
Replies: 36
Views: 2003

Re: Stockfish with and without contempt

I really don't think it can be quite as bad as you make it out to be (if used reasonably). The eval of any engine is already uncertain (other than hard cutoffs). So if one has uncertain nodes that eval as draw, they could just as well not be. Which is why SF's evaluate.cpp has received many patches ...
by jhellis3
Wed Nov 06, 2019 2:34 am
Forum: Computer Chess Club: General Topics
Topic: My predicted 3rd Gen threadripper prices
Replies: 41
Views: 1863

Re: My predicted 3rd Gen threadripper prices

I would guess they will want something at the $999 price point, so I would guess that for the lowest end, $1199 max. Top tier, I'd go $2999 max but I'm less certain of that. The 28 core skylake-x is $2978, so yeah I'd go upper bound of $2999 but likely less.... maybe $2699? As for site reliability w...
by jhellis3
Tue Nov 05, 2019 9:44 pm
Forum: Computer Chess Club: General Topics
Topic: Hard pawn endgame
Replies: 11
Views: 805

Re: Hard pawn endgame

This is difficult. The best problems solver = Crystal is clueless after 10 minutes. It don't reach enough depth I think. It is a good test position. Unfortunately, Crystal has no mechanisms in place to deal appropriately with such positions. However, it is pretty clear how one might address the iss...
by jhellis3
Tue Nov 05, 2019 3:48 pm
Forum: Computer Chess Club: General Topics
Topic: LCZero blunders in TCEC cup 4
Replies: 20
Views: 2081

Re: LCZero blunders in TCEC cup 4

It depends on what input features the NNs have access to and how they are trained. For A0, it was certainly possible that such an effect occurred (due to the presence of movecount). For LC0, the likelihood & potential impact is much less as it does not have movecount as an input. It does still have ...
by jhellis3
Thu Oct 31, 2019 9:39 pm
Forum: Computer Chess Club: General Topics
Topic: Tactics cannot be very important for chess
Replies: 90
Views: 4592

Re: Tactics cannot be very important for chess

Dann Corbit wrote:Adding points to the existing score is a stupid way to solve it.
This is actually a very funny statement :D .
by jhellis3
Wed Oct 30, 2019 7:13 pm
Forum: Computer Chess Club: General Topics
Topic: Tactics cannot be very important for chess
Replies: 90
Views: 4592

Re: Tactics cannot be very important for chess

I would say NNs have the *most* tactical potential due to superior move ordering & eval. The issues present in the most prevalent networks are more the result of search, training method(s), and input parameters (lack thereof) than any inherent deficit in NNs themselves. Just my 2cp anyway.....
by jhellis3
Wed Oct 30, 2019 2:11 am
Forum: Computer Chess Club: General Topics
Topic: "Fat fritz" stronger than Stockfish ?
Replies: 79
Views: 13590

Re: "Fat fritz" stronger than Stockfish ?

It isn't like 42850 is a very high bar.... I'm sure the Chinese already have private nets +200 or 300 Elo....
by jhellis3
Mon Oct 28, 2019 8:02 pm
Forum: Computer Chess Club: General Topics
Topic: Some words on Leela Go Zero
Replies: 4
Views: 959

Re: Some words on Leela Go Zero

What you are experiencing is likely due to a couple of reasons... Initial search space for Go is massive, and NNs are (relatively) slow to eval, which means resolving tactics of any significant depth by search is rather hopeless. One might try something similar to what I use in Chess for fortress de...
by jhellis3
Mon Oct 14, 2019 8:36 pm
Forum: Computer Chess Club: General Topics
Topic: Winning after Qf6+!
Replies: 97
Views: 9269

Re: Winning after Qf6+!

Yeah, did some more manual analysis and it looks lost regardless at that point.... it might be possible for Black to still play Nxa3 after the bishop captures h5 with the same Nb5 plan, but it makes no (apparent) sense to play this way unless allowing the pawn capture with the rook creates some issu...