endgame: passed pawns vs. pieces

[jdart]

Interesting endgame in an online game, Arasan-crafty:

[d] 4r3/8/p1k3K1/1p6/2p5/2R3B1/5P2/8 w - -

here Arasan played Re3 and lost. It should probably "know" that passed pawns are generally stronger with fewer pieces on the board, but seeing that White is tactically lost after Re3 Rxe3 is harder.

In addition to avoiding Re3 in the above position an engine should play Rxe3 when the exchange is offered:

[d] 4r3/8/p1k3K1/1p6/2p5/4R1B1/5P2/8 b - -

Rybka 3 plays it at shallow depths, switches to another move and then sees a win at ply 18:

yes 11 280 +28 131296 Rxe3 fxe3 c3 Be5 c2 Bb2 Kd5 Kf5 a5 e4+
Kc5 e5
yes 12 450 +17 222992 Rxe3 fxe3 c3 Be5 c2 Bb2 Kd5 Kf5 a5 e4+
Kc5 e5 a4
yes 13 720 +0 349880 Rxe3 fxe3 c3 Be5 c2 Bb2 Kd5 Kf5 a5 e4+
Kc6 e5 a4 Ke6
no 13 5720 +0 1522416 Rc8 Re6+ Kd5 Rd6+ Ke4 Re6+ Kd5 Rd6+ Ke4
Re6+ Kd5 Rd6+ Ke4 Re6+
no 14 14940 +4 3658120 Rc8 Re6+ Kc5 Rxa6 c3 Be5 c2 Bb2 b4 Kf5
Re8 Kf4 b3 Ra7
no 15 23080 +5 5740000 Rc8 Re6+ Kc5 Rxa6 c3 Be5 c2 Bb2 b4 Kf5
Kb5 Ra1 Re8 Kf4
no 16 43500 +0 10636312 Rc8 f4 b4 Be1 Kb5 f5 c3 f6 c2 Bd2 Rd8
Re5+ Kc6 Re6+
no 17 72560 +0 16768472 Rc8 f4 b4 Be1 b3 Bc3 Rg8+ Kf5 Rg2 Re1 Rc2
Re6+ Kd5 Re5+
no 18 90660 +0 20892872 Rc8 f4 b4 Be1 b3 Bc3 Rg8+ Kf5 Rg2 Re1 Rc2
Re6+ Kd5 Re5+
yes 18 125310 +699 26344240 Rxe3 fxe3 b4 e4 b3 Be5 a5 Kf7 a4 Bb2 c3
Bxc3 a3 e5
yes 19 136220 +717 29052304 Rxe3 fxe3 b4 e4 b3 Be5 a5 Kf7 a4 Bb2 c3
Bxc3 a3 e5
yes 20 143410 +717 30618128 Rxe3 fxe3 b4 e4 b3 Be5 a5 Kf7 a4 Bb2 c3
Bxc3 a3 e5

Shredder does better here. Crafty 21.6 also plays Rxe3 right away and doesn't vary:

11-> 0.12 -1.48 1. ... Rxe3 2. fxe3 c3 3. Be5 c2 4.
Bb2 b4 5. Kf5 a5 6. e4 a4
12 0.14 -1.16 1. ... Rxe3 2. fxe3 c3 3. Be5 c2 4.
Bb2 b4 5. Kf5 a5 6. e4 b3 7. e5
12-> 0.45 -1.16 1. ... Rxe3 2. fxe3 c3 3. Be5 c2 4.
Bb2 b4 5. Kf5 a5 6. e4 b3 7. e5
13 0.47 -1 1. ... Rxe3!! (1.8Mnps)
13 0.48 -1.58 1. ... Rxe3 2. fxe3 c3 3. Be5 c2 4.
Bb2 b4 5. Kf5 a5 6. e4 b3 7. e5 a4
13-> 0.83 -1.58 1. ... Rxe3 2. fxe3 c3 3. Be5 c2 4.
Bb2 b4 5. Kf5 a5 6. e4 b3 7. e5 a4
14 0.93 -1.37 1. ... Rxe3 2. fxe3 c3 3. Be5 c2 4.
Bb2 b4 5. Kf5 a5 6. e4 b3 7. e5 a4
8. e6
14-> 2.56 -1.37 1. ... Rxe3 2. fxe3 c3 3. Be5 c2 4.
Bb2 b4 5. Kf5 a5 6. e4 b3 7. e5 a4
8. e6
15 2.64 -1 1. ... Rxe3!! (1.9Mnps)
15 2.76 -1.65 1. ... Rxe3 2. fxe3 c3 3. Be1 c2 4.
Bd2 b4 5. Kf5 b3 6. Bc1 Kc5 7. e4 a5
8. e5 a4
15-> 5.06 -1.65 1. ... Rxe3 2. fxe3 c3 3. Be1 c2 4.
Bd2 b4 5. Kf5 b3 6. Bc1 Kc5 7. e4 a5
8. e5 a4
16 5.23 -1.56 1. ... Rxe3 2. fxe3 c3 3. Be1 c2 4.
Bd2 b4 5. Kf5 b3 6. Bc1 Kc5 7. e4 a5
8. e5 a4 9. e6
16-> 14.40 -1.56 1. ... Rxe3 2. fxe3 c3 3. Be1 c2 4.
Bd2 b4 5. Kf5 b3 6. Bc1 Kc5 7. e4 a5
8. e5 a4 9. e6
17 14.95 -1 1. ... Rxe3!! (1.9Mnps)
17 15.79 -2.74 1. ... Rxe3 2. fxe3 a5 3. Kf5 b4 4.
Be1 c3 5. Ke4 c2 6. Bd2 a4 7. Kd3 b3
8. Bc1 Kd5 9. e4+ Ke5 10. Bb2+ Kf4
<HT>
17-> 25.50 -2.74 1. ... Rxe3 2. fxe3 a5 3. Kf5 b4 4.
Be1 c3 5. Ke4 c2 6. Bd2 a4 7. Kd3 b3
8. Bc1 Kd5 9. e4+ Ke5 10. Bb2+ Kf4
<HT>
18 25.89 +1 1. ... Rxe3 (1.8Mnps)
18 26.33 -1 1. ... Rxe3!! (1.8Mnps)
18 27.01 -3.94 1. ... Rxe3 2. fxe3 b4 3. Kf6 a5 4.
e4 b3 5. Be5 a4 6. Bb2 c3 7. Bxc3 a3
8. Bb4 b2 9. Bxa3 b1=Q 10. e5

--Jon

[George Tsavdaris]

[d] 4r3/8/p1k3K1/1p6/2p5/4R1B1/5P2/8 b - -

Rybka 3 plays it at shallow depths, switches to another move and then sees a win at ply 18:

yes 11 280 +28 131296 Rxe3 fxe3 c3 Be5 c2 Bb2 Kd5 Kf5 a5 e4+
Kc5 e5
yes 12 450 +17 222992 Rxe3 fxe3 c3 Be5 c2 Bb2 Kd5 Kf5 a5 e4+
Kc5 e5 a4
yes 13 720 +0 349880 Rxe3 fxe3 c3 Be5 c2 Bb2 Kd5 Kf5 a5 e4+
Kc6 e5 a4 Ke6
no 13 5720 +0 1522416 Rc8 Re6+ Kd5 Rd6+ Ke4 Re6+ Kd5 Rd6+ Ke4
Re6+ Kd5 Rd6+ Ke4 Re6+
no 14 14940 +4 3658120 Rc8 Re6+ Kc5 Rxa6 c3 Be5 c2 Bb2 b4 Kf5
Re8 Kf4 b3 Ra7
no 15 23080 +5 5740000 Rc8 Re6+ Kc5 Rxa6 c3 Be5 c2 Bb2 b4 Kf5
Kb5 Ra1 Re8 Kf4
no 16 43500 +0 10636312 Rc8 f4 b4 Be1 Kb5 f5 c3 f6 c2 Bd2 Rd8
Re5+ Kc6 Re6+
no 17 72560 +0 16768472 Rc8 f4 b4 Be1 b3 Bc3 Rg8+ Kf5 Rg2 Re1 Rc2
Re6+ Kd5 Re5+
no 18 90660 +0 20892872 Rc8 f4 b4 Be1 b3 Bc3 Rg8+ Kf5 Rg2 Re1 Rc2
Re6+ Kd5 Re5+
yes 18 125310 +699 26344240 Rxe3 fxe3 b4 e4 b3 Be5 a5 Kf7 a4 Bb2 c3
Bxc3 a3 e5
yes 19 136220 +717 29052304 Rxe3 fxe3 b4 e4 b3 Be5 a5 Kf7 a4 Bb2 c3
Bxc3 a3 e5
yes 20 143410 +717 30618128 Rxe3 fxe3 b4 e4 b3 Be5 a5 Kf7 a4 Bb2 c3
Bxc3 a3 e5

I wonder how you get these results from Rybka 3.
My Rybka 3 default 32-bit SP on one CPU(a slow one) prefers this move immediately and with a big score for black from the beginning:

Analysis by Rybka 3 1-cpu 32-bit :

1...Re8xe3 2.f2xe3 b5-b4 3.Bg3-e1 a6-a5 4.Kg6-f5 c4-c3 5.Kf5-e4 c3-c2 6.Be1-d2
-+ (-2.50) Depth: 9 00:00:00 19kN
1...Re8xe3
-+ (-2.70) Depth: 10 00:00:00 30kN
1...Re8xe3
-+ (-2.90) Depth: 10 00:00:00 37kN
1...Re8xe3
-+ (-3.30) Depth: 10 00:00:01 48kN
1...Re8xe3
-+ (-4.10) Depth: 10 00:00:01 94kN, tb=1
1...Re8xe3 2.f2xe3 b5-b4 3.Bg3-e5 c4-c3 4.Kg6-f5 Kc6-c5 5.Kf5-e4 c3-c2 6.Be5-b2 Kc5-c4 7.Ke4-f3 Kc4-d3 8.e3-e4 a6-a5 9.Bb2-c1 a5-a4 10.e4-e5
-+ (-4.11) Depth: 10 00:00:03 191kN, tb=3
1...Re8xe3 2.f2xe3 b5-b4 3.Bg3-e5 c4-c3 4.Kg6-f5
-+ (-4.11) Depth: 11 00:00:03 199kN, tb=4
1...Re8xe3 2.f2xe3 b5-b4 3.Bg3-e5 c4-c3 4.Kg6-f5 Kc6-c5
-+ (-4.11) Depth: 12 00:00:04 225kN, tb=10
1...Re8xe3
-+ (-4.31) Depth: 13 00:00:05 279kN, tb=21
1...Re8xe3
-+ (-4.51) Depth: 13 00:00:06 350kN, tb=28
1...Re8xe3
-+ (-4.91) Depth: 13 00:00:07 433kN, tb=34
1...Re8xe3
-+ (-5.58) Depth: 13 00:00:42 2627kN, tb=381
1...Re8xe3
-+ (-6.17) Depth: 14 00:11:05 48382kN, tb=1527


Also i wonder at which hardware you've made this analysis and how much time did Rybka 3 made to reach 15,16,17 and at the 18 ply.

[jdart]

My mistake: I had my polyglot ini file set up for Rybka 1.2f, not Rybka 3!

It is as you say: Rybka 3 plays it rapidly and with increasing score. 1.2f has a problem but not Rybka 3.

Re hw: I am running it on my Athlon 64x2 4400+ with 256MB hash.

--Jon

[Eelco de Groot]

Ancalagon would never willingly give up the blocking square c3 and the control of the Rook over a3 and b3 also helps to decrease the value of three - three -passed pawns by just a tiny amount. But the value of these three pawns is so high in Ancalagon's eyes that he would much rather call it a day, if the GUI allows it. Re3? Never!

Is there any move here left for White that would justify not resigning immediately? I think even Rybka 3 would need contempt of at least the value of a queen not to resign or to display 0.00? I don't think that Ancalagon evaluates the given position correctly but I do think White is in some trouble here.

[d]4r3/8/p1k3K1/1p6/2p5/2R3B1/5P2/8 w - -

Engine: Ancalagon 1.08 Testversion 002 (256 MB)
by Tord Romstad

8.00 0:00 -9.62 1.f4 Rg8+ 2.Kf5 Kd5 3.Rxc4 bxc4
4.Be1 Ra8 5.Ba5 Rf8+ 6.Kg5 (70.956) 188

9.00 0:00 -9.43 1.f4 Rg8+ 2.Kf6 Rxg3 3.Rxg3 b4 4.Ke5 b3
5.f5 Kd7 6.Rg7+ Ke8 7.Kd4 (322.663) 337

9.00 0:01 -9.17 1.Kf7 Re4 2.f3 Rd4 3.Be5 Rd7+ 4.Ke6 a5
5.f4 Rd8 6.Kf5 (434.454) 360

10.01 0:01 -10.29 1.Kf7 Re4 2.f3 Rd4 3.Be5 Kd5 4.Kf6 Rd3
5.Kf5 a5 6.Kf4 Rxc3 (612.135) 383

10.02 0:02 -7.13 1.f4 Rg8+ 2.Kh7 Rxg3 3.Rxg3 Kd7 4.f5 Ke8
5.Rc3 b4 6.Rxc4 b3 7.Kg6 (925.157) 401

11.01 0:03 -7.86 1.f4 Rg8+ 2.Kh7 Rf8 3.Kg7 Rxf4
4.Bxf4 a5 5.Kg6 Kd5 6.Kf5 b4 7.Rxc4 Kxc4 (1.441.288) 410

12.01 0:10 -7.62 1.f4 a5 2.f5 Rg8+ 3.Kf7 Rg4 4.f6 b4
5.Be5 a4 6.Ke6 bxc3 7.Bxc3 a3 8.f7 a2
9.Ke5 Rg1 10.f8Q Rg5+ 11.Kd4 (4.420.651) 438

13.01 0:23 -10.33 1.f4 Kd5 2.f5 b4 3.Rxc4 Kxc4 4.Bd6 Rg8+
5.Kf7 b3 6.f6 Kd5 7.Ba3 Rg2 8.Ke7 Re2+
9.Kf7 a5 (9.807.791) 426

14.01 0:37 -9.47 1.f4 Kd5 2.f5 b4 3.Rxc4 Kxc4 4.Bd6 Rg8+
5.Kf7 b3 6.f6 Kd5 7.Ba3 Rg2 8.Ke7 Re2+
9.Kf8 Ra2 10.Bb4 (16.575.791) 436

15.01 0:52 -9.50 1.f4 Kd5 2.f5 b4 3.Rxc4 Kxc4 4.Bd6 Rg8+
5.Kf7 b3 6.f6 Kd5 7.Ba3 Rg2 8.Ke7 Re2+
9.Kf8 Ra2 10.f7 Rxa3 11.Ke7 (23.212.143) 442

16.01 1:22 -10.82 1.f4 Kd5 2.f5 b4 3.Rxc4 Kxc4 4.Bd6 Rg8+
5.Kf7 Rg2 6.f6 b3 7.Ba3 Kd5 8.Ke7 Re2+
9.Kf7 Ra2 10.Kg6 Rxa3 11.Kf5 (37.205.091) 452

17.01 5:38 -10.76 1.f4 Kd5 2.f5 b4 3.Rf3 b3 4.f6 b2
5.Rf5+ Kc6 6.f7 Rf8 7.Rf6+ Kb7 8.Be5 b1Q+
9.Kg5 Rxf7 10.Rxf7+ Kc6 11.Rf4 Kd5
12.Rf5 Ke6 13.Rf4 Qg1+ 14.Kh5 (152.073.917) 448

18.01 8:33 -10.70 1.f4 Kd5 2.f5 b4 3.Rf3 b3 4.f6 b2
5.Rf5+ Kc6 6.Kg7 Re3 7.Bf2 b1Q 8.Ra5 Rf3
9.Rxa6+ Kd5 10.Ra5+ Ke6 11.Ra6+ Ke5
12.Rb6 Qxb6 13.Bxb6 Rxf6 (227.996.039) 443

19.01 16:11 -10.82 1.f4 Kd5 2.f5 b4 3.Rf3 b3 4.f6 b2
5.Rf5+ Kc6 6.Kg7 Re3 7.Bh4 b1Q 8.Rf4 c3
9.f7 Qg1+ 10.Kf8 Kd5 11.Rf5+ Kd6
12.Rg5 Qf1 13.Bg3+ Kd7 14.Rd5+ (421.352.853) 433

best move: f2-f4 time: 29:42.281 min n/s: 436.757 nodes: 780.360.000

[jdart]

I agree White is probably lost even without playing Re3, but if you're getting scores like -9 in a shallow search then IMO you are probably valuing the passed pawns too highly.

--jon

[Eelco de Groot]

Okay so maybe resigning would not yet be necessary in the position, I see Naum 3.1 just thinks it is a draw. Can't be right! But maybe the passed pawn eval actually does need a revision, somewhere . I just am not yet certain where exactly.. Just scaling it down when the opponent has more pieces to throw against the pawns might help..

[Event "Shoot-out"]
[Site "The Dragon's Den"]
[Date "2008.08.17"]
[Round "?"]
[White "Naum 3.1"]
[Black "Naum 3.1"]
[Result "1/2-1/2"]
[SetUp "1"]
[FEN "4r3/8/p1k3K1/1p6/2p5/2R3B1/5P2/8 w - -"]

1. Kf5 Kd5 2. Rc1 Rf8+ 3. Kg4 b4 4. Rd1+ Kc6 5. f4 c3
6. Rc1 Kd5 7. f5 Rc8 8. Rd1+ Kc4 9. Be5 c2 10. Ra1 a5
11. f6 a4 12. Bb2 a3 13. Bc1 Rc7 14. Kf5 Kd5 15. Kg6 Rc6
16. Kf5 Rc7 17. Kg6 Rc6 18. Kf5 Rc7 1/2-1/2

[Eelco de Groot]

I agree White is probably lost even without playing Re3, but if you're getting scores like -9 in a shallow search then IMO you are probably valuing the passed pawns too highly.

--jon

Jon I think you are absolutely right. There may also some bugs in passed pawn eval somewhere, that may make the situation worse. And I think hereby I will apologize in advance to the CCC community for the sometimes below par analysis because of new program changes, it will probably happen often but I'm trying to make it better!

Eelco

[Eelco de Groot]

[Event "Shoot-out"]
[Site "The Dragon's Den"]
[Date "2008.08.17"]
[Round "?"]
[White "Naum 3.1"]
[Black "Naum 3.1"]
[Result "1/2-1/2"]
[SetUp "1"]
[FEN "4r3/8/p1k3K1/1p6/2p5/2R3B1/5P2/8 w - -"]

1. Kf5 Kd5 2. Rc1 Rf8+ 3. Kg4 b4 4. Rd1+ Kc6 5. f4 c3
6. Rc1 Kd5 7. f5 Rc8 8. Rd1+ Kc4 9. Be5 c2 10. Ra1 a5
11. f6 a4 12. Bb2 a3 13. Bc1 Rc7 14. Kf5 Kd5 15. Kg6 Rc6
16. Kf5 Rc7 17. Kg6 Rc6 18. Kf5 Rc7 1/2-1/2

But a counterexample is also possible, Ancalagon wins against Naum!


[Event "6 Minutes/Game + 3 Seconds/Move"]
[Site "Engine Match"]
[Date "2008.08.17"]
[Round "1"]
[White "Naum 3.1"]
[Black "Ancalagon 1.08 Testversion 002"]
[Result "0-1"]
[SetUp "1"]
[FEN "4r3/8/p1k3K1/1p6/2p5/2R3B1/5P2/8 w - -"]

1. Kf5 {0.00/13 6s} Kd5 {+0.10/15 15s eval by Naum here
from previous shoot-out game} 2. Ra3 {-0.13/15 19s
(Rc1)} b4 {+14.50/13 14s} 3. Ra5+ {0.00/13 5s} Kd4
{+13.94/14 14s} 4. Be5+ {0.00/16 14s (Rxa6)}
{[d]4r3/8/p7/R3BK2/1ppk4/8/5P2/8 b - -

Without hesitation, Ancalagon sacs the exchange:}

4...Rxe5+ {!?!}

{+11.13/13 10s} 5. Rxe5 {0.00/15 9s} c3 {+11.19/13 9s}
6. Re4+ {-4.08/19 8s} Kc5 {+11.39/11 1s} 7. Re8 {0.00/21
11s (Re5+)} b3 {+9.58/13 11s} 8. Rc8+ {0.00/22 1:02m (Ke4)}
Kd4 {+6.74/15 11s} 9. Rd8+ {-4.96/22 43s} Kc4 {+6.62/12 0s}
10. Ke4 {-4.96/22 19s} b2 {+6.86/15 13s} 11. Rc8+ {-4.96/22
44s} Kb3 {+7.08/16 12s} 12. Kd4 {-10.26/21 38s} Ka2
{+7.41/13 12s} 13. Rc6 {-7.73/17 4s (Rxc3)} a5 {+8.60/13
9s} 14. Rxc3 {-M38/19 32s} b1=Q {+10.47/12 11s} 15. f4
{-M37/8 0s} Qb6+ {+11.70/12 11s} 16. Kd5 {-M36/7 0s} Qb7+
{+14.90/14 11s} 17. Rc6 {-M35/6 0s (Kd4)} Qd7+ {+18.68/13
9s} 18. Kc5 {-M32/14 0s} Qf5+ {+19.00/14 10s} 19. Kb6
{-M31/7 0s} a4 {+18.43/12 8s} 20. Kb7 {-M20/13 3s} a3
{+22.90/13 10s} 21. Rc7 {-M19/8 0s (Ka6)} Kb3 {+39.03/10
9s} 22. Kb8 {-M15/9 1s (Kc6)} a2 {+86.61/8 0s} 23. Rb7+
{-M14/7 0s} Kc4 {+M12/14 21s} 24. Ra7 {-M13/4 0s (Rc7+)}
Qxf4+ {+M9/12 1s} 25. Kb7 {-M12/3 0s (Kc8)} Qf7+ {+M7/9 0s}
26. Kb6 {-M6/1 0s} Qxa7+ {+M6/8 0s} 27. Kxa7 {-M6/1 0s} Kc5
{+M5/8 0s} 28. Ka6 {-M5/1 0s (Kb8)} a1=Q+ {+M4/6 0s}
29. Kb7 {-M4/1 0s} Qg7+ {+M3/6 0s} 30. Ka6 {-M3/1 0s (Kc8)}
Qd7 {+M2/6 0s} 31. Ka5 {-M2/1 0s} Qb5# {+M1/6 0s} 0-1


[d]8/8/8/Kqk5/8/8/8/8 w - -

[pijl]

Interesting endgame in an online game, Arasan-crafty:

[d] 4r3/8/p1k3K1/1p6/2p5/2R3B1/5P2/8 w - -

here Arasan played Re3 and lost. It should probably "know" that passed pawns are generally stronger with fewer pieces on the board, but seeing that White is tactically lost after Re3 Rxe3 is harder.

CTD030 on a slow P4-2.4 does not consider Re3 at any point in time:

Code: Select all

ply   score   time           nodes pv
 7&   0.668   0:00.10        37558 g6f7 c6d7 f2f4 e8e7 f7f6 e7e6 f6f5 a6a5 g3f2
 7.   0.668   0:00.17        56212 g6f7 c6d7 f2f4 e8e7 f7f6 e7e6 f6f5 a6a5 g3f2
 8&   0.329   0:00.25        83358 g6f7 e8e2 f2f4 c6c5 c3a3 c5b4 a3a6 c4c3
 8.   0.329   0:00.67       261078 g6f7 e8e2 f2f4 c6c5 c3a3 c5b4 a3a6 c4c3
 9&   0.643   0:00.85       340392 g6f7 e8a8 f2f4 c6c5 f4f5 b5b4 c3e3 c5d4 g3f4
 9.   0.643   0:00.96       384346 g6f7 e8a8 f2f4 c6c5 f4f5 b5b4 c3e3 c5d4 g3f4
10&   0.439   0:01.29       528184 g6f7 e8e2 f2f4 c6d5 c3a3 b5b4 a3a5 d5d4 f4f5 e2e3 g3d6
10.   0.439   0:02.03       826060 g6f7 e8e2 f2f4 c6d5 c3a3 b5b4 a3a5 d5d4 f4f5 e2e3 g3d6
11&   0.463   0:03.81      1564885 g6f7 e8a8 f2f4 c6d5 f4f5 a8a7 f7g6 b5b4 c3c1 c4c3
11.   0.463   0:04.95      2098335 g6f7 e8a8 f2f4 c6d5 f4f5 a8a7 f7g6 b5b4 c3c1 c4c3
12&  -0.057   0:06.76      2916205 g6f7 e8e2 f2f4 c6c5 f4f5 b5b4 c3f3 b4b3 f5f6 b3b2 f3f5 c5d4 f5f1
12&   0.332   0:16.31      6937681 f2f4 c6d5 g6f7 e8e2 f4f5
12.   0.332   0:18.15      7766522 f2f4 c6d5 g6f7 e8e2 f4f5
13&  -1.154   0:25.28     10654279 f2f4 b5b4 c3c4 c6b5 c4c7 e8g8 g6f7 g8g3 f4f5 g3h3 f7g7 h3c3 c7c3 b4c3 f5f6
13&   0.015   0:30.23     12895764 g6f7 e8e2 f2f4 c6c5 f4f5 b5b4 c3c1 c4c3 f5f6 c3c2 f7f8 b4b3 f6f7
13.   0.015   0:47.07     20451669 g6f7 e8e2 f2f4 c6c5 f4f5 b5b4 c3c1 c4c3 f5f6 c3c2 f7f8 b4b3 f6f7
14&  -0.535   1:00.73     26753444 g6f7 e8e2 c3c1 c6d5 c1d1 d5e4 d1c1 a6a5 f2f4 a5a4 f7f6 a4a3 f4f5 a3a2 c1a1
14&  -0.459   1:26.60     38443225 g6f6 c6d5 c3c1 e8c8 g3e5 c8f8 f6e7 f8f2 e5c3 f2e2 e7f6 e2e3 c1d1 e3d3 d1c1 d3f3 f6e7
14&  -0.283   2:11.07     58333547 c3c2 c6c5 g6f7 e8c8 f2f4 b5b4 f4f5 c4c3 f5f6 c5d4 c2h2
14.  -0.283   3:53.62    103647730 c3c2 c6c5 g6f7 e8c8 f2f4 b5b4 f4f5 c4c3 f5f6 c5d4 c2h2
15&  -0.239   4:14.18    112794658 c3c2 c6c5 g6f7 e8c8 f2f4 b5b4 f4f5 c4c3 c2e2 c5d4 g3f2 d4d3 e2e8 c8c4 e8d8 d3c2 f5f6
15.  -0.239   7:32.39    196121323 c3c2 c6c5 g6f7 e8c8 f2f4 b5b4 f4f5 c4c3 c2e2 c5d4 g3f2 d4d3 e2e8 c8c4 e8d8 d3c2 f5f6
16&  -0.546   8:37.20    225552004 c3c2 c6c5 g6f7 e8c8 g3e5 b5b4 f7e7 b4b3 c2c1 a6a5 f2f4 a5a4 f4f5 c5d5 e5b2 c8g8
16&  -0.202   9:38.10    249819567 g6f7 e8e2 c3c1 c6d5 f2f4 e2e3 g3e1 e3f3 c1d1 d5e4 f7e6 e4f4 d1d4 f4e3 e1c3 a6a5 e6e5 a5a4
16.  -0.202  12:35.75    318831501 g6f7 e8e2 c3c1 c6d5 f2f4 e2e3 g3e1 e3f3 c1d1 d5e4 f7e6 e4f4 d1d4 f4e3 e1c3 a6a5 e6e5 a5a4
17&  -0.454  13:29.21    338727952 g6f7 e8e2 c3c1 c6d5 f2f4 e2e3 g3h4 e3f3 h4f6 b5b4 c1d1 f3d3 d1b1 b4b3 f4f5 a6a5 f7g7 a5a4

In addition to avoiding Re3 in the above position an engine should play Rxe3 when the exchange is offered:

[d] 4r3/8/p1k3K1/1p6/2p5/4R1B1/5P2/8 b - -

Started with it, briefly selected Rc8 due to temporary drop in the score, but came back with at +1 increasing to +2 and +4 score:

Code: Select all

ply   score   time           nodes pv
10&   1.570   0:00.12        78382 e8e3 f2e3 c4c3 g3e5 c3c2 e5b2 b5b4 g6f5 b4b3 f5e4
10.   1.570   0:00.14        82046 e8e3 f2e3 c4c3 g3e5 c3c2 e5b2 b5b4 g6f5 b4b3 f5e4
11&   1.846   0:00.21       131968 e8e3 f2e3 c4c3 e3e4 b5b4 g3e5 c3c2 e5f4 c6c5 f4e3 c5c4 g6f5
11.   1.846   0:00.25       140382 e8e3 f2e3 c4c3 e3e4 b5b4 g3e5 c3c2 e5f4 c6c5 f4e3 c5c4 g6f5
12&   0.579   0:00.46       287358 e8e3 f2e3 c4c3 e3e4 b5b4 g3e5 c6c5 g6f7 a6a5 e5f4 c5d4 e4e5
12.   0.579   0:00.53       313492 e8e3 f2e3 c4c3 e3e4 b5b4 g3e5 c6c5 g6f7 a6a5 e5f4 c5d4 e4e5
13&   0.000   0:00.87       546916 e8e3 f2e3 c4c3 e3e4 c3c2 g3f4 b5b4 g6f7 b4b3 f4c1 c6d6 c1f4 d6c6
13&   0.433   1:17.96     31228290 e8c8 f2f4 b5b4 g3e1 c6b5 e3e5 b5a4 f4f5 c4c3 f5f6 b4b3 e1c3 c8c3 f6f7
13.   0.433   1:18.00     31240828 e8c8 f2f4 b5b4 g3e1 c6b5 e3e5 b5a4 f4f5 c4c3 f5f6 b4b3 e1c3 c8c3 f6f7
14&  -0.034   1:32.92     36888980 e8c8 f2f4 b5b4 g3e1 b4b3 e1c3 c8b8 f4f5 b3b2 c3b2 b8b2 f5f6 c6d5 g6f7
14&   1.866   1:33.53     37272713 e8e3 f2e3 c4c3 e3e4 b5b4 g3e1 c3c2 e1d2 b4b3 d2c1 c6c5 e4e5 c5d5 g6f5
14.   1.866   1:33.60     37308877 e8e3 f2e3 c4c3 e3e4 b5b4 g3e1 c3c2 e1d2 b4b3 d2c1 c6c5 e4e5 c5d5 g6f5
15&   1.512   1:34.39     37853377 e8e3 f2e3 c4c3 e3e4 b5b4 g3e1 c3c2 e1d2 b4b3 d2c1 c6d6 g6f5 d6c5 e4e5 c5b5
15.   1.512   1:34.56     37957476 e8e3 f2e3 c4c3 e3e4 b5b4 g3e1 c3c2 e1d2 b4b3 d2c1 c6d6 g6f5 d6c5 e4e5 c5b5
16&   1.216   1:36.00     38959123 e8e3 f2e3 c4c3 e3e4 b5b4 g3e1 c3c2 e1d2 c6c5 e4e5 b4b3 d2c1 c5d5 g6f6 a6a5 e5e6
16.   1.216   1:56.45     46777440 e8e3 f2e3 c4c3 e3e4 b5b4 g3e1 c3c2 e1d2 c6c5 e4e5 b4b3 d2c1 c5d5 g6f6 a6a5 e5e6
17&   2.171   2:09.21     56296765 e8e3 f2e3 b5b4 g3e1 a6a5 g6f5 c4c3 f5e4 c3c2 e1d2 a5a4 e4d3 b4b3 d2c1
17.   2.171   2:09.87     56601402 e8e3 f2e3 b5b4 g3e1 a6a5 g6f5 c4c3 f5e4 c3c2 e1d2 a5a4 e4d3 b4b3 d2c1
18&   4.290   2:18.60     63428552 e8e3 f2e3 b5b4 e3e4 a6a5 g6f7 b4b3 g3e5 a5a4 e5b2 c4c3 b2c3 a4a3 c3b4 b3b2 b4a3 b2b1q e4e5
18.   4.290   2:19.79     64133070 e8e3 f2e3 b5b4 e3e4 a6a5 g6f7 b4b3 g3e5 a5a4 e5b2 c4c3 b2c3 a4a3 c3b4 b3b2 b4a3 b2b1q e4e5

Richard.

[jdart]

[d] 4r3/8/p1k3K1/1p6/2p5/2R3B1/5P2/8 w - -

Rybka prefers Bh4 here which at least allows White to delay a bit. For example, 1. Bh4 Kc5 2. Kf7 Re2 3. Ra3 (not 3. Be7+ Rxe7+ Kxe7, which
loses as your other exchange sac line) b4 4. Ra5+ and White can at least win the a pawn.