Stuck

[Lyudmil Tsvetkov]

I am so stuck, really, currently.

I meant to write that it is probably time for a steak, but I wrote 'stake' instead.

Maybe it is also time to bet with Arjun that an uniform 60cps bonus for a blocked pawn on the 6th rank, excluding only h6 that already has a big bonus, will pass both short and LTC. I.e., a uniform bonus only for blocked pawns on the 6th, excluding the 5th rank, where term clashes are bigger and chances to go astray increase.

[Lyudmil Tsvetkov]

There is no doubt that SF underestimates minors vs pawns in the middlegame, but that would be true only of the middlegame.

So you are saying a minor piece should be worth more than 4.2 pawns?

I am saying that, in SF code, a minor piece, only in the middlegame, and only within the specific minor-pawns imbalance should be worth more than 4.2 for the B and more than 4.1 for the N. But a minor piece within N vs B or 2 minors + pawns vs Q, or minor vs R is a different thing.

The correction should be valid only when SF has minor for 3 enemy pawns, something like Reuven tried for the R vs minors, where I think it failed at LTC because of the quadratic stuff.

[jhellis3]

Code: Select all

Stockfish 300414 64 SSE4.2 by Tord Romstad, Marco Costalba and Joona Kiiski
position fen 1rbr3k/6bp/p5p1/2pN4/2Pp1PPq/3Q4/P2B2B1/3R1RK1 b - - 0 33
eval
           Eval term |    White    |    Black    |    Total    
                     |   MG    EG  |   MG    EG  |   MG    EG  
---------------------+-------------+-------------+-------------
Material, PST, Tempo |   ---   --- |   ---   --- |  2.97  2.94 
  Material imbalance |   ---   --- |   ---   --- |  0.14  0.14 
               Pawns |   ---   --- |   ---   --- | -0.17 -0.15 
             Knights |  0.20  0.20 |  0.00  0.00 |  0.20  0.20 
             Bishops | -0.12 -0.19 | -0.09 -0.23 | -0.03  0.05 
               Rooks |  0.07  0.04 |  0.17  0.08 | -0.09 -0.04 
              Queens |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 
            Mobility |  0.68  1.38 |  0.46  1.08 |  0.22  0.30 
         King safety | -0.40 -0.19 |  0.25 -0.06 | -0.64 -0.12 
             Threats |  0.00  0.00 |  0.22  0.48 | -0.22 -0.48 
        Passed pawns |  0.00  0.00 |  0.34  0.26 | -0.34 -0.26 
               Space |  0.17  0.00 |  0.17  0.00 |  0.00  0.00 
---------------------+-------------+-------------+-------------
               Total |   ---   --- |   ---   --- |  1.95  2.56 

Total Evaluation: 2.02 (white side)

position fen 1r1r3k/6bp/p5p1/2pN4/2Pp1Pbq/3Q4/P2B2B1/3R1RK1 w - - 0 34
eval
           Eval term |    White    |    Black    |    Total    
                     |   MG    EG  |   MG    EG  |   MG    EG  
---------------------+-------------+-------------+-------------
Material, PST, Tempo |   ---   --- |   ---   --- |  2.01  1.81 
  Material imbalance |   ---   --- |   ---   --- |  0.12  0.12 
               Pawns |   ---   --- |   ---   --- | -0.42 -0.59 
             Knights |  0.20  0.20 |  0.00  0.00 |  0.20  0.20 
             Bishops | -0.09 -0.14 | -0.05 -0.19 | -0.04  0.05 
               Rooks |  0.07  0.04 |  0.17  0.08 | -0.09 -0.04 
              Queens |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 
            Mobility |  0.65  1.31 |  0.74  1.72 | -0.09 -0.41 
         King safety | -0.79 -0.19 |  0.22 -0.06 | -1.02 -0.12 
             Threats |  0.00  0.00 |  0.16  0.39 | -0.16 -0.39 
        Passed pawns |  0.00  0.00 |  0.34  0.29 | -0.34 -0.29 
               Space |  0.17  0.00 |  0.19  0.00 | -0.02  0.00 
---------------------+-------------+-------------+-------------
               Total |   ---   --- |   ---   --- |  0.28  0.52 

Total Evaluation: 0.31 (white side)

Ok, the above evaluations are for the position after White recaptures the knight, and after Black recaptures the pawn with the bishop. In both cases, you can see there is nothing wrong with the material eval. SF considers White to have to have a 2 pawn advantage there. The actual "problem" if there is one is the king safety bringing down the score. Perhaps, we might reducing the king safety weight if behind in material, and possibly pawn weight if behind in non-pawn material (though you can see this is not as large a factor as king safety).

[Lyudmil Tsvetkov]

Code: Select all

Stockfish 300414 64 SSE4.2 by Tord Romstad, Marco Costalba and Joona Kiiski
position fen 1rbr3k/6bp/p5p1/2pN4/2Pp1PPq/3Q4/P2B2B1/3R1RK1 b - - 0 33
eval
           Eval term |    White    |    Black    |    Total    
                     |   MG    EG  |   MG    EG  |   MG    EG  
---------------------+-------------+-------------+-------------
Material, PST, Tempo |   ---   --- |   ---   --- |  2.97  2.94 
  Material imbalance |   ---   --- |   ---   --- |  0.14  0.14 
               Pawns |   ---   --- |   ---   --- | -0.17 -0.15 
             Knights |  0.20  0.20 |  0.00  0.00 |  0.20  0.20 
             Bishops | -0.12 -0.19 | -0.09 -0.23 | -0.03  0.05 
               Rooks |  0.07  0.04 |  0.17  0.08 | -0.09 -0.04 
              Queens |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 
            Mobility |  0.68  1.38 |  0.46  1.08 |  0.22  0.30 
         King safety | -0.40 -0.19 |  0.25 -0.06 | -0.64 -0.12 
             Threats |  0.00  0.00 |  0.22  0.48 | -0.22 -0.48 
        Passed pawns |  0.00  0.00 |  0.34  0.26 | -0.34 -0.26 
               Space |  0.17  0.00 |  0.17  0.00 |  0.00  0.00 
---------------------+-------------+-------------+-------------
               Total |   ---   --- |   ---   --- |  1.95  2.56 

Total Evaluation: 2.02 (white side)

position fen 1r1r3k/6bp/p5p1/2pN4/2Pp1Pbq/3Q4/P2B2B1/3R1RK1 w - - 0 34
eval
           Eval term |    White    |    Black    |    Total    
                     |   MG    EG  |   MG    EG  |   MG    EG  
---------------------+-------------+-------------+-------------
Material, PST, Tempo |   ---   --- |   ---   --- |  2.01  1.81 
  Material imbalance |   ---   --- |   ---   --- |  0.12  0.12 
               Pawns |   ---   --- |   ---   --- | -0.42 -0.59 
             Knights |  0.20  0.20 |  0.00  0.00 |  0.20  0.20 
             Bishops | -0.09 -0.14 | -0.05 -0.19 | -0.04  0.05 
               Rooks |  0.07  0.04 |  0.17  0.08 | -0.09 -0.04 
              Queens |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 
            Mobility |  0.65  1.31 |  0.74  1.72 | -0.09 -0.41 
         King safety | -0.79 -0.19 |  0.22 -0.06 | -1.02 -0.12 
             Threats |  0.00  0.00 |  0.16  0.39 | -0.16 -0.39 
        Passed pawns |  0.00  0.00 |  0.34  0.29 | -0.34 -0.29 
               Space |  0.17  0.00 |  0.19  0.00 | -0.02  0.00 
---------------------+-------------+-------------+-------------
               Total |   ---   --- |   ---   --- |  0.28  0.52 

Total Evaluation: 0.31 (white side)

Ok, the above evaluations are for the position after White recaptures the knight, and after Black recaptures the pawn with the bishop. In both cases, you can see there is nothing wrong with the material eval. SF considers White to have to have a 2 pawn advantage there. The actual "problem" if there is one is the king safety bringing down the score. Perhaps, we might reducing the king safety weight if behind in material, and possibly pawn weight if behind in non-pawn material (though you can see this is not as large a factor as king safety).

Thanks Joseph.

Good remarks.

I think the excellent idea is to reduce king safety weight if ahead in non-pawn material (that is what counts, pawns do not count here when it is about king safety), as you should reduce the white king safety weight, and white is in material advantage, non-pawn.

Also very good reducing pawn weight when behind in non-pawn material.

But both should be valid only for the middlegame.

What about reducing mobility when behind in non-pawn material?

[Lyudmil Tsvetkov]

[d]6k1/5p2/1p1p1Pp1/1PpP2Pp/2P4P/8/8/6K1 w - - 0 1
So, I would scale here the eval by just 10%.

[d]6k1/5p2/1p3Pp1/pPp3Pp/P1P4P/8/8/6K1 w - - 0 1
But not here, as 2 of the central or semicentral files are open.

[d]6k1/5p2/1p3Pp1/pPp3Pp/P1Pp3P/3P4/8/6K1 w - - 0 1
I would scale here eval by 20-30%.

[Lyudmil Tsvetkov]

Again on this topic, which concerns many engines, and especially SF, too much.

Below my refined suggestions how to implement scaling down of score with opposite-coloured bishops, so that it gives a boost to elo, and not vice-versa:

- scale for the entire game, and not just for the endgame (the Berlin is an excellent example that you should scale for the entire game, and not just the endgame; if I have time, I might post some game where SF is the black protagonist in the Berlin and, believe me, SF is notorious for mishandling the Berlin)
- never scale with queens on the board, regardless of whether it is middlegame or endgame. It is just dangerous, even in the endgame, as opposite-coloured bishops are known to specifically foster king attacks, and that is especially true when queens are on. In most cases a king attack with queens on might not materialise, but what to do if it materialises?
- scale down some 10% if more than half of possible non-pawn material is still on the board, 20% if less than 1/3 of material is still present, 30% and more when material is very low; with just a single bishop each side, you can scale 50%.

I think those are not very bad rules that might be successfully implemented, if one wishes too. They could also bring a nice portion of elo.

What do you think of engine eval in such positions? People post diagrams where engines are wrong all too often, and still engines in general persist in their eval, no one cares to improve opposite-colour bishops play, even Komodo.

[arjuntemurnikar]

Maybe it is also time to bet with Arjun that an uniform 60cps bonus for a blocked pawn on the 6th rank, excluding only h6 that already has a big bonus, will pass both short and LTC. I.e., a uniform bonus only for blocked pawns on the 6th, excluding the 5th rank, where term clashes are bigger and chances to go astray increase.

Ok, I have pushed two quick tests to the framework to see if it is any improvement. 60 seems very high, so I have also pushed 40.

[mcostalba]

- scale for the entire game, and not just for the endgame (the Berlin is an excellent example that you should scale for the entire game, and not just the endgame; if I have time, I might post some game where SF is the black protagonist in the Berlin and, believe me, SF is notorious for mishandling the Berlin)

I have submitted a test to do it.

[Lyudmil Tsvetkov]

- scale for the entire game, and not just for the endgame (the Berlin is an excellent example that you should scale for the entire game, and not just the endgame; if I have time, I might post some game where SF is the black protagonist in the Berlin and, believe me, SF is notorious for mishandling the Berlin)

I have submitted a test to do it.

Many thanks, Marco!

I do not know when I will be able to repay all the good things SF is doing for me.

Please, bear in mind though, that smmoth transition is important, and the condition that scaling is done only when queens are off is possibly even more important.

Many thanks again.
I know you will soon dig my grave to the bottom when another couple of 'supposedly good ideas' gets smashed to dust and sand into the framework.

[Lyudmil Tsvetkov]

Maybe it is also time to bet with Arjun that an uniform 60cps bonus for a blocked pawn on the 6th rank, excluding only h6 that already has a big bonus, will pass both short and LTC. I.e., a uniform bonus only for blocked pawns on the 6th, excluding the 5th rank, where term clashes are bigger and chances to go astray increase.

Ok, I have pushed two quick tests to the framework to see if it is any improvement. 60 seems very high, so I have also pushed 40.

Thanks Arjun.

You are my biggest benefactor and possibly grave-digger. (or do they say undertaker?)

Concerning values, Stefan's 260 was even more shocking, but you have to consider that 60cps SF values is some 30cps standard ones, so that 30cps standard bonus for 6th rank is not big at all. The question is if it will still clash with something...apart from the already assigned bonus for edge h6/a6 storming pawn, which should be negligeable overall. (but excluding just the h6/a6 storming pawns from this would be even more consistent). I would be very grateful if you leave at least one of the tests for STC regardless of how the values score at 5 seconds. You know, those pawns scale differently, you have the bitter storm experience.

Many thanks again, Arjun!