Don't know how to make engine understand that this position is lost for white.
Should I assign a huge bonus to the knight on f4. But that would be different if white had a knight on g3.
[d] r1b2rk1/p4ppp/2p2q2/4p3/2BpPn2/5N2/PPP2PPP/R3QRK1 w - - 2 16
Evaluation
Moderators: hgm, Rebel, chrisw
-
- Posts: 881
- Joined: Sun Dec 27, 2020 2:40 am
- Location: Bremen, Germany
- Full name: Thomas Jahn
Re: Evaluation
Chess.com says the position is -178cp on depth 37. MinimalChess says it's -149cp on depth 12. I think that's close enough...
MinimalChess evaluation is based exclusively on 12 PSTs + 1 extra table that gives tiny bonuses/maluses based on what other squares & pieces each piece can see. All 13 tables are fully autotuned.
So if you haven't tried auto-tuning yet give it a try. You just need to come up with something that can encode chess knowledge but then you don't have to encode the chessknowledge yourself. Great for programmers that can't play chess themselves (like me) but also great for the others!
MinimalChess evaluation is based exclusively on 12 PSTs + 1 extra table that gives tiny bonuses/maluses based on what other squares & pieces each piece can see. All 13 tables are fully autotuned.
So if you haven't tried auto-tuning yet give it a try. You just need to come up with something that can encode chess knowledge but then you don't have to encode the chessknowledge yourself. Great for programmers that can't play chess themselves (like me) but also great for the others!
-
- Posts: 1871
- Joined: Sat Nov 25, 2017 2:28 pm
- Location: France
Re: Evaluation
Minic says around -300 at depth > 20, seems quite ok.
(static eval -77)
(static eval -77)
Last edited by xr_a_y on Mon Jul 05, 2021 1:24 pm, edited 2 times in total.
-
- Posts: 27796
- Joined: Fri Mar 10, 2006 10:06 am
- Location: Amsterdam
- Full name: H G Muller
Re: Evaluation
It is not clear whether your question refers to the static evaluation, or to the search score. I suppose to the former, as a little bit of search here should surely show that things end badly for white. And a question about search scores would be pretty meaningless anyway without specifying a search depth. With deep-enough search you would see anything.
I guess this is a King-Safety issue. As a Chess player the proximity of the Queen while g2 is already attacked would kindle concern. Presumably a neural net would also have learned this as a dangerous pattern: the King crammed into a corner, cut of by its own Rook (which itself cannot usefully participate in the defense) from any aid, with a Knight already attacking the Pawn shield, a Bishop attacking the square in front of one of the Pawns in this shield, and a Queen somewhere in the open area before the shield.
I guess this is a King-Safety issue. As a Chess player the proximity of the Queen while g2 is already attacked would kindle concern. Presumably a neural net would also have learned this as a dangerous pattern: the King crammed into a corner, cut of by its own Rook (which itself cannot usefully participate in the defense) from any aid, with a Knight already attacking the Pawn shield, a Bishop attacking the square in front of one of the Pawns in this shield, and a Queen somewhere in the open area before the shield.
-
- Posts: 881
- Joined: Sun Dec 27, 2020 2:40 am
- Location: Bremen, Germany
- Full name: Thomas Jahn
Re: Evaluation
The static eval is -70 for MMC (-67 from the PSTs and -3 from the thirteenth table) and that's probably not "bad enough" to do the position justice but the patterns HGM describes need more than a few tables to encode.
What does your eval think about this, Henk?
-
- Posts: 7218
- Joined: Mon May 27, 2013 10:31 am
Re: Evaluation
I changed some values and I now get -0.53.
But my kingsafety currently only counts pawnshield.
It adds a bonus to Nf4, bonus to bishop/queen attacking kingzone.
But probably should multiply these factors which I do not.
Hi hi hi. Pawn on f2 is bad for kingsafety (in this case)
But my kingsafety currently only counts pawnshield.
It adds a bonus to Nf4, bonus to bishop/queen attacking kingzone.
But probably should multiply these factors which I do not.
Hi hi hi. Pawn on f2 is bad for kingsafety (in this case)
-
- Posts: 334
- Joined: Sat Feb 25, 2012 10:42 pm
- Location: Stockholm
Re: Evaluation
The big problem for white is that knight on f3 can only move safely to one other square. The rook on f1 can not move at all. The black pawn on d4 is a potential passer. Another small inconvenience with the f3 knight is that it is in front of its pawn thus hindering its advancement.
-
- Posts: 7218
- Joined: Mon May 27, 2013 10:31 am
Re: Evaluation
Hmm there is a rule in chess that knight on fifth row is worth a rook.
Probably that only holds in optimal circumstances.
In this case Nf4 could be chased away by g3 or exchanged for a knight in the long term.
Probably that only holds in optimal circumstances.
In this case Nf4 could be chased away by g3 or exchanged for a knight in the long term.
-
- Posts: 334
- Joined: Sat Feb 25, 2012 10:42 pm
- Location: Stockholm
Re: Evaluation
Yes but chasing the knight away will lead to bad pawn structure. The knight on f4 is extra good because it is close to the opponent’s king and there are plenty of material left for check mate.
-
- Posts: 373
- Joined: Wed Mar 22, 2006 10:17 am
- Location: Novi Sad, Serbia
- Full name: Karlo Balla
Re: Evaluation
Tuning the engine against a particular position is (almost) always a bad idea...
Moreover, this position is tactical, even if it does not look like that (Ng2->Bh3->Qf3...)
Moreover, this position is tactical, even if it does not look like that (Ng2->Bh3->Qf3...)
Best Regards,
Karlo Balla Jr.
Karlo Balla Jr.