interesting position for eval

[jdart]

from an online game:

[d] 5nrk/1q3npp/p7/4pN1P/2p1p1Q1/4B3/2P3P1/4R1K1 w - -

Black is ahead 3 pawns but his Rook is buried, some of the pawns are weak and White has an attack - so what is the net eval?

Rybka 3 with a moderate depth (16) search likes Ng3 here with a score of about -0.75 pawn for White.

Arasan rates the material much more highly than the positional factors, with a score below -2 for White.

Crafty is in the middle with a score about -1.8 at depth 16.

--Jon

[swami]

Bright CCT Depth 18

-1.28

Prefers Ng3.

[Eelco de Groot]

The position is far from quiet so the evaluations may not have very much to do with the static evaluation of the position. Both Rybka 3 and Bright are known to look very deep so I think that explains they see a higher eval for White. Black's pawns are weak especially e4 and e5 are probably going to fall.

Some difficulties for eval in no particular order:

♥ What do you do with all those open files if you have no Rooks or the one Rook you have is locked up.
♣ Knights are not terribly good at protecting plus pawns especially if they are all isolated.
♠ Several passed pawns but isolated and none of them threatening anything, no Rooks to support them.
♦ White has some compensation for the material in better placed pieces for an attack on the black King position. But not enough pieces that it would end in mate so he has to use this to convert to a more equal endgame. This probably means you have to scale down the value for attacked squares etc. around the King at some point, but when exactly.. I find that a difficult thing to judge, in Ancalagon I still evaluate King Safety also in the endgame part of the score but I'm not sure that it is not by far too much. I have not looked at that recently.

With Ancalagon I see two consecutive fail highs in the last iterations so that was another sign for my 'evaluation' that tactics play a big role in the position, it is maybe better to look at the evaldifferences when search has had time to find out most of the tactics?


[d]5nrk/1q3npp/p7/4pN1P/2p1p1Q1/4B3/2P3P1/4R1K1 w - -

Engine: Ancalagon 1.3 WS180 Build 171 (Athlon 2009 MHz, 256 MB)
by Romstad, Costalba, Kiiski, de Groot

2.00 0:00 -2.62 1.Rd1 Ne6 2.Nd6 (5.204) 17

2.00 0:00 -2.45 1.Rf1 Ne6 (6.745) 21

2.00 0:00 -1.72 1.Bd2 Qb6+ 2.Kf1 (42.659) 113

2.00 0:00 -1.54 1.Bf2 Ne6 2.Rxe4 (43.937) 117

3.00 0:00 -1.56 1.Bf2 Ne6 2.Rxe4 Qb1+ 3.Be1 (79.169) 181

4.00 0:00 -1.72 1.Bf2 Ne6 2.Rxe4 Qb1+ 3.Be1 a5 (286.719) 353

4.00 0:00 -1.41 1.Ng3 Qd5 2.Nxe4 Ne6 (307.450) 364

5.00 0:00 -1.47 1.Ng3 Nd7 2.Nxe4 Nh6 3.Bxh6 gxh6 (388.416) 400

6.01 0:01 -1.78 1.Ng3 Nd7 2.Qxe4 Qb4 3.Rd1 Nf6 4.Qf5 (637.172) 468

7.01 0:03 -1.80 1.Ng3 Nd7 2.Qxe4 Qb4 3.Rd1 Nd6 4.Qd5 (1.816.214) 543

8.01 0:11 -1.92 1.Ng3 Nd7 2.Qxe4 Qb4 3.Rd1 Rd8 4.Qf5 Nd6 (6.550.235) 588

8.02 0:11 -1.91 1.Bd2 (6.908.718) 589

8.11 0:42 -1.91 1.Bf2 (23.972.777) 568

9.01 0:48 -1.74 1.Bf2 Ne6 2.Qxe4 Qxe4 3.Rxe4 Rc8
4.Kf1 a5 5.Ne7 Rc7 6.Nd5 (28.050.163) 579

10.01 2:43 -1.78 1.Bf2 Ne6 2.Nd6 Nxd6 3.Qxe6 Qc6
4.Qxe5 Re8 5.Qc5 Qxc5 6.Bxc5 Nf5 (99.518.346) 608

10.02 2:46 -1.66 1.Rd1 Ne6 2.Nd6 Qe7 3.Nxc4 Rc8 4.Nb6 Rxc2
5.Qxe4 Re2 (101.377.819) 610

11.01 4:33 -1.94 1.Rd1 Ne6 2.Nd6 Nxd6 3.Qxe6 Ne8
4.Rd8 Qb1+ 5.Kh2 Qb5 6.Rd5 Nc7
7.Rxb5 Nxe6 8.Rxe5 (170.465.125) 622

11.02 4:51 -1.70 1.Bf2 Ne6 2.Qxe4 Qxe4 3.Rxe4 Rc8
4.Be1 Neg5 5.Re3 Rd8 6.Ra3 Nd6
7.Nxd6 Rxd6 (180.950.265) 620

11.05 6:44 -1.70 1.Ng3 (249.748.147) 618

12.01 12:00 -1.50 1.Ng3 (439.795.261) 610

13.01 15:51 -1.47 1.Ng3 Nd6 2.Rd1 Qc7 3.Nxe4 Nxe4
4.Qxe4 Nd7 5.Qf3 Nf6 6.h6 Rf8
7.hxg7+ Qxg7 8.Rf1 Nd7 (575.654.686) 604

best move: Nf5-g3 time: 47:57.141 min n/s: 572.900 nodes: 1.648.260.241

Eelco

[jdart]

I don't think there's a shallow tactical shot here. So it is interesting that Rybka even from low depths sees a lot of compensation for White. Not clear why. Maybe a large mobility bonus. Maybe a big pawn structure component. Rybka's score implies it would sac two pawns (at least) to get this position for White. I'm not a GM so it's hard for me to say if that's accurate or not.

[michiguel]

I don't think there's a shallow tactical shot here. So it is interesting that Rybka even from low depths sees a lot of compensation for White. Not clear why. Maybe a large mobility bonus. Maybe a big pawn structure component. Rybka's score implies it would sac two pawns (at least) to get this position for White. I'm not a GM so it's hard for me to say if that's accurate or not.

Gaviota scores oscilate from -0.98 at ply one to -1.15 at ply 15. Mobility is a big factor in current Gaviota's eval. I also do not think that tactics are an issue here.

Miguel

[michiguel]

I don't think there's a shallow tactical shot here. So it is interesting that Rybka even from low depths sees a lot of compensation for White. Not clear why. Maybe a large mobility bonus. Maybe a big pawn structure component. Rybka's score implies it would sac two pawns (at least) to get this position for White. I'm not a GM so it's hard for me to say if that's accurate or not.

Gaviota scores oscilate from -0.98 at ply one to -1.15 at ply 15. Mobility is a big factor in current Gaviota's eval. I also do not think that tactics are an issue here.

Miguel

After ply 16 the score is -1.07 and the PV:
1.Rf1 a5 2.Ng3 Nd6 3.Rd1 Qc6 4.Qg5 c3 5.Qxe5 Nc4 6.Qxc3 Ne6 7.h6 Qc7 8.Nxe4 Nxe3 9.Qxe3 Qxc2

The static score is -1.35

Miguel

[diep]

from an online game:

[d] 5nrk/1q3npp/p7/4pN1P/2p1p1Q1/4B3/2P3P1/4R1K1 w - -

Black is ahead 3 pawns but his Rook is buried, some of the pawns are weak and White has an attack - so what is the net eval?

Rybka 3 with a moderate depth (16) search likes Ng3 here with a score of about -0.75 pawn for White.

Arasan rates the material much more highly than the positional factors, with a score below -2 for White.

Crafty is in the middle with a score about -1.8 at depth 16.

--Jon

A problem of modern computerchess is that a position like this is playable with black. Diep might rate it at -0.553, yet you simply lose with white when there is no 'shot' at a specific time.

00:00 200 2 0 (4) 1 (0,0) -5.317 Nf5xg7 Rg8xg7
00:00 300 3 0 (4) 1 (0,0) -1.264 Re1-e2
00:00 400 4 0 (4) 1 (0,0) -0.706 Re1-d1
00:00 500 5 0 (4) 1 (0,0) -0.326 Be3-c5
00:00 700 7 0 (4) 1 (0,0) -0.141 Re1-f1
00:00 11700 117 0 (4) 2 (0,1) -0.486 Re1-f1 Nf7-h6
00:00 46400 464 0 (4) 3 (1,5) -0.385 Re1-f1 Nf8-e6 c2-c3
00:00 74433 2233 0 (4) 4 (4,15) -0.942 Re1-f1 Nf8-d7 Qg4-h4 Rg8-f8
00:00 108037 8643 0 (4) 5 (11,50) -0.764 Re1-f1 Nf8-d7 h5-h6 g7xh6 Qg4-h4
00:00 174578 24441 0 (4) 6 (16,123) -1.039 Re1-f1 g7-g6 h5xg6 Nf8xg6 Qg4-h3 Rg8-f8
++ g4-e2 procnr=3 terug=-1038 org=[-1039;-1038] newwindow=[-1039;520000]
00:00 219469 50478 0 (4) 6 (30,225) -0.976 Qg4-e2 Qb7-b4 Re1-a1 a6-a5 Qe2-g4 a5-a4 Qg4xe4
00:00 271132 92185 0 (4) 7 (39,423) -0.757 Qg4-e2 Nf8-d7 Qe2xc4 Nd7-f6 Re1-a1 Rg8-c8 Qc4-e2
++ e1-f1 procnr=1 terug=-756 org=[-757;-756] newwindow=[-757;520000]
00:00 285132 105499 0 (4) 7 (42,471) -0.533 Re1-f1 g7-g6 Nf5-g3 g6xh5 Qg4-h4 Nf7-d6 Ng3xh5
00:00 328410 190478 0 (4) 8 (62,764) -0.616 Re1-f1 Nf8-e6 Nf5-h6 Nf7xh6 Qg4xe6 Qb7-b2 Qe6xc4 Nh6-g4
00:00 407490 330067 0 (4) 9 (76,1092) -0.504 Re1-f1 a6-a5 Qg4-e2 g7-g6 h5xg6 Rg8xg6 Nf5-h6 Nf7xh6 Rf1xf8 Kh8-g7
00:01 483401 671928 0 (4) 10 (140,1851) -0.432 Re1-f1 Nf8-e6 Nf5-h6 Nf7xh6 Qg4xe6 Qb7-c8 Qe6-d5 Nh6-g4 Be3-c5 Ng4-f6 Qd5xc4
00:02 540705 1373392 0 (4) 11 (242,3021) -0.526 Re1-f1 Nf8-e6 Nf5-d6 Nf7xd6 Qg4xe6 Nd6-e8 h5-h6 a6-a5 Kg1-h2 Qb7-b5 Rf1-f7 a5-a4 Qe6-f5
00:06 616806 3848871 0 (4) 12 (441,5022) -0.773 Re1-f1 a6-a5 Qg4-e2 Qb7-c7 Qe2-g4 Nf8-d7 Nf5-e7 Nd7-f6 Rf1xf6 Qc7xe7 Rf6-a6 Qe7-b4 Qg4xe4 Qb4-e1 Kg1-h2
++ g4-e2 procnr=1 terug=-772 org=[-773;-772] newwindow=[-773;520000]
00:13 647995 8670185 0 (4) 13 (717,8004) -0.773 Re1-f1 a6-a5 Qg4-e2 Qb7-c7 Qe2-g4 Nf8-d7 Nf5-e7 Nd7-f6 Rf1xf6 Qc7xe7 Rf6-a6 Qe7-b4 Qg4xe4 Qb4-e1 Kg1-h2
00:25 667241 17234860 0 (4) 14 (1168,12742) -0.553 Re1-f1 a6-a5 Qg4-h4 Nf8-e6 Nf5-e7 Rg8-f8 Be3-c5 Nf7-h6 Ne7-g6 h7xg6 Bc5xf8 Nh6-f5 Qh4-h3 Qb7-b6 Kg1-h2 Qb6-e3 Bf8-b4
00:39 689771 27149409 0 (4) 15 (1442,15983) -0.553 Re1-f1 a6-a5 Qg4-h4 Nf8-e6 Nf5-e7 Rg8-f8 Be3-c5 Nf7-h6 Ne7-g6 h7xg6 Bc5xf8 Nh6-f5 Qh4-h3 Qb7-b6 Kg1-h2 Qb6-e3