Rook+pawns vs two minors

Discussion of chess software programming and technical issues.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
metax
Posts: 344
Joined: Wed Sep 23, 2009 3:56 pm
Location: Germany

Re: Rook+pawns vs two minors

Post by metax » Mon Jan 18, 2010 2:41 pm

Sven Schüle wrote: Which values do you mean by "the values would imply ..." here? If you mean minor=3.75/rook=5.50 then my calculator says 2 x 3.75 - 5.50 = 2, not 2.5. So probably you mean something else which I can't see yet.
I meant what you also meant, but I have miscalculated this. :oops: :)
Of course you are right, and 2*3.75 - 5.50 is 2 pawns. Nevertheless the question is still if this is too much (I don't think so) or rather too little.

Sven Schüle wrote:As to your question: while traditional chess theory told us that R+P nearly equals two minors (except in case of a bishop pair), today it seems clear that it is closer to R+2P at least. To be honest: I don't know what the value should be but as a chessplayer I believe that values close to three pawns are highly exaggerated *for human chess play*. However, it may turn out to be quite different for engines. While parameter tuning for one engine may deliver good results with R+3P vs. two minors, another one may tend to more "human-like" R+2P or even less and get best results this way round.
The 'traditional material count' every beginner is teached first when he starts learning chess and which is pretty common among chess players < 2000 Elo (and also in not few amateur engines) although it estimates many material configurations very wrong:

{ P=1; N=3; B=3; R=5; Q=9 }

Here is N=B=3P and R+P=N+B.

In practical chess play I have seen many cases (also in my own games) where a rook and one or two pawns were traded for two minors, and I have always seen the two minors win. In some cases things like bishop pair or rook pair redundancy were involved, but not in all.

Sven Schüle wrote:I guess there is no "general answer" since the positional evaluation is so different between engines, which makes it nearly impossible in my opinion to compare only the pure material value constants of two engines.

I just made a small experiment with Rybka 2.2n2 (no R3 available here :-( ): from the initial position I removed Nb1, Bc1 and Ra8, Pa7 and started a very quick analysis. The PV (e4 g6 Nf3 Bg7 c3 Nc6 Qa4 Nf6 e5 Nd5) showed a value of -0.95 after few seconds. Then I also removed Pc7, which gave -0.51 (Nf3 Nf6 e3 Nc6 c3 g6 d4 e6 Bd3 Bg7 O-O). Finally I removed Pd7 and the PV was "e3 Nc6 Bb5 Nf6 Nf3 e6 a4 Bd7 a5 Qb8 Nd4 Bd6" with a value of +0.01. So R+3P gave about an equal play, of course many positional factors add into this here. But interesting was that R+1P was not -2.00 but gave a PV with only about -1.00.

It is not representative, and it is also about search, not only about material value constants, I know. I just found it interesting.

Sven
I think Rybka is able to evaluate material imbalances more exact here as in the imbalance table, every plausible case seems to be covered. Things like these can not be evaluated very exactly.

metax
Posts: 344
Joined: Wed Sep 23, 2009 3:56 pm
Location: Germany

Re: Rook+pawns vs two minors

Post by metax » Mon Jan 18, 2010 2:56 pm

I have continued your experiment a little bit. I built up the positions you mentioned and let some engines analyse it for 30 seconds. Here are my results:

[D]1nbqkbnr/1ppppppp/8/8/8/8/PPPPPPPP/R2QKBNR w KQk - 0 1

Code: Select all

1nbqkbnr/1ppppppp/8/8/8/8/PPPPPPPP/R2QKBNR w KQk -

Engine&#58; Rybka 2.2n2 mp 32-bit &#40;128 MB&#41;
von Vasik Rajlich

11.01  0&#58;02   -0.90    1.e3 g6 2.Sf3 Lg7 3.c3 Sf6 4.Ld3 Sc6 &#40;273.834&#41; 132 

12.01  0&#58;04   -0.98    1.e3 g6 2.Lb5 Lg7 3.d4 Sc6 4.Sf3 Sf6 
                       5.O-O &#40;619.513&#41; 132 

12.10  0&#58;10   -0.90    1.e4 Sc6 2.Sf3 Sf6 3.e5 Sd5 4.Lc4 Sb6 
                       5.De2 &#40;1.121.667&#41; 112 

13.01  0&#58;14   -0.90    1.e4 Sc6 2.Sf3 e6 3.Lb5 Sf6 4.De2 Lc5 
                       5.O-O-O O-O &#40;1.449.196&#41; 102 

Bester Zug&#58; e2-e4 Zeit&#58; 0&#58;30.202 min  K/s&#58; 2.813.928  Knoten&#58; 2.813.928

Code: Select all

1nbqkbnr/1ppppppp/8/8/8/8/PPPPPPPP/R2QKBNR w KQk -

Engine&#58; Stockfish 1.6 JA &#40;128 MB&#41;
von Tord Romstad, Marco Costalba, Joona Kii

15.01  0&#58;01   -1.33    1.Sf3 Sf6 2.e3 e6 3.d4 d5 4.Ld3 Ld6 
                       5.O-O O-O 6.c4 dxc4 7.Lxc4 Sc6 8.Ld3 &#40;1.234.108&#41; 670 

16.01  0&#58;03   -1.37    1.Sf3 Sf6 2.e3 e6 3.d4 d5 4.Ld3 Ld6 
                       5.O-O O-O 6.c4 dxc4 7.Lxc4 Sc6 8.Ld3 e5 
                       9.Lb5 e4 10.Lxc6 bxc6 &#40;2.566.854&#41; 679 

16.02  0&#58;16   -1.21    1.e4 d5 2.e5 Sh6 3.Sf3 g6 4.Lb5+ c6 
                       5.Le2 c5 6.O-O Sc6 7.Lb5 Lg7 8.c4 Sg4 
                       9.d4 cxd4 10.Sxd4 Lxe5 11.Sxc6 bxc6 
                       12.Lxc6+ Kf8 13.Dxd5 Dxd5 14.Lxd5 &#40;11.006.795&#41; 650 

Bester Zug&#58; e2-e4 Zeit&#58; 0&#58;30.061 min  K/s&#58; 643.364  Knoten&#58; 19.340.172 

Code: Select all

1nbqkbnr/1ppppppp/8/8/8/8/PPPPPPPP/R2QKBNR w KQk -

Engine&#58; Twisted Logic 20090922 &#40;128 MB&#41;
von Edsel Apostol, Philippines

13/27  0&#58;02   -0.92    1.e3 Sf6 2.d4 d5 3.Lb5+ c6 4.Ld3 Db6 
                       5.Sf3 Dxb2 6.O-O Sbd7 7.Db1 Dxb1 
                       8.Tfxb1 &#40;969.274&#41; 456 

14/28  0&#58;08   -0.85    1.e3 Sf6 2.d4 d5 3.c3 Sc6 4.Sf3 Se4 
                       5.Lb5 Ld7 6.O-O e6 7.Sd2 Dh4 8.Sxe4 Dxe4 &#40;3.853.151&#41; 430 

15/29  0&#58;23   -0.93    1.e3 e5 2.Lb5 Sc6 3.Se2 Dg5 4.O-O Sf6 
                       5.c4 Lb4 6.f4 exf4 7.a3 Lc5 8.Sxf4 &#40;9.665.521&#41; 415 

15/29  0&#58;28   -0.76    1.d4 d5 2.Sf3 Sc6 3.e3 Dd6 4.c3 e5 
                       5.dxe5 Sxe5 6.Da4+ c6 7.O-O-O Sxf3 
                       8.gxf3 Sf6 9.Ld3 Le7 &#40;11.809.952&#41; 413 

Bester Zug&#58; d2-d4 Zeit&#58; 0&#58;30.155 min  K/s&#58; 413.412  Knoten&#58; 12.427.607 
They all think black is better off, that is pretty obvious. But by different amounts: Rybka thinks it is -0.9, Stockfish favors black by 1.2 and TL likes the white position best with -0.76.

zamar
Posts: 613
Joined: Sun Jan 18, 2009 6:03 am

Re: Rook+pawns vs two minors

Post by zamar » Mon Jan 18, 2010 3:14 pm

metax wrote: They all think black is better off, that is pretty obvious. But by different amounts: Rybka thinks it is -0.9, Stockfish favors black by 1.2 and TL likes the white position best with -0.76.
Sorry to complicate this further, but unfortunately "pawn value" is very relative concept :(

Do we use value of the edge-pawn or center-pawn?
Do we use value of middle-game pawn or end-game pawn?

Fx. Stockfish's scores are usually somewhat or even radically higher than scores shown by other engines...
Joona Kiiski

metax
Posts: 344
Joined: Wed Sep 23, 2009 3:56 pm
Location: Germany

Re: Rook+pawns vs two minors

Post by metax » Mon Jan 18, 2010 3:34 pm

zamar wrote:Sorry to complicate this further, but unfortunately "pawn value" is very relative concept :(

Do we use value of the edge-pawn or center-pawn?
Do we use value of middle-game pawn or end-game pawn?
I think it makes sense to normalize the pawn value to 1.00. Some engines have reduced the value of a pawn to 0.70 or 0.80 in some game phases, but I think this only makes things more complicated.
My opinion is that the modification of the value of pieces based on their position does not make sense because the pieces' material values are only supposed to give an average of their value. The positional refinements based on the position of the pieces should in my opinion rather be done using piece-square tables or other evaluation components. Adjusting the value of a piece does only make things even more complicated.

zamar wrote:Fx. Stockfish's scores are usually somewhat or even radically higher than scores shown by other engines...
I know this. Part of the reason for this may be the aggressive and active playing style of Stockfish.
ChessMind's evaluation for the above position is -1.50, however. :shock:
I think I still have to do some adjustment of the material values...

jdart
Posts: 3842
Joined: Fri Mar 10, 2006 4:23 am
Location: http://www.arasanchess.org

Re: Rook+pawns vs two minors

Post by jdart » Mon Jan 18, 2010 6:30 pm

Arasan gives a small bonus for the minor side in R vs. B or RR vs RB - so it is valuing the exchange at a bit less than 2 pawns. It will play Rxf7 here. And its score for the Gruenfeld position at ply 12 is similar to what Stockfish has. So it's doing something roughly comparable to other programs.

bm Rxf7
0 0.00 Rxf7 +2.03 4
1 0.00 Rxf7 +2.03 83
2 0.00 -- +1.53 1007
2 0.00 Rxf7 +0.10 1182
2 0.00 a4 +0.27 1700
2 0.02 e5 +0.28 2158
2 0.02 Nd5 +0.59 2942
3 0.02 Nd5 +0.51 4378
3 0.02 Rxf7 +0.52 5266
4 0.02 Rxf7 +0.14 8037
4 0.03 Nd5 +0.55 12890
5 0.05 Nd5 +0.67 37307
6 0.14 Nd5 +0.55 105692
7 0.20 Nd5 +0.55 181542
8 0.45 Nd5 +0.47 509213
9 1.02 Nd5 +0.43 1529796
10 1.47 Nd5 +0.31 2336565
11 3.63 Nd5 +0.01 6386591
11 4.44 e5 +0.14 8172918
12 9.42 e5 +0.20 16489477
13 13.59 e5 +0.21 23700919
13 16.63 Rxf7 +0.42 29347434
14 19.16 Rxf7 +0.52 34212972
15 24.83 Rxf7 +0.72 46069589
16 34.09 Rxf7 +0.54 65276878

Post Reply