SF did a mistake in TCEC game 26

[Nay Lin Tun]

[d]1rbr2k1/2qn1pb1/p1pp2p1/1p2p1Pp/PP1PP3/2P1BP2/3QBNP1/R4RK1 b - - 0 17

In this position, SF played 17...bxa4 ?? and lost the game! After that move, white has a protected passed pawn b4 supported by c3 while black has a very weak backward a6 pawn! I can't understand why bxa4 was played.

[cc2150dx]

well, here's a general idea what komodo might have played.

FEN: 1rbr2k1/2qn1pb1/p1p3p1/1p1pp1Pp/PP1PP3/2P1BP2/3QBNP1/R4RK1 w - - 0 18

Komodo-10.2-64-bit:

28 00:49 352,264k 7,103k -0.24 d6-d5 e4xd5 Nd7-b6 d5-d6 Qc7xd6 Nf2-e4 Qd6-f8 a4xb5 a6xb5 Ne4-f6+ Bg7xf6 g5xf6 Nb6-d5 Be3-f2 Nd5-f4 Qd2-e3 Bc8-f5 Rf1-e1 Nf4-d5 Qe3-d2 Nd5xf6 Be2-d1 Nf6-d5 Bd1-b3 Bf5-e6 Qd2-g5 Nd5-c7 Qg5xe5 Be6xb3 Qe5xc7 Rb8-c8 Qc7-e5 Rd8-d5 Qe5-f6
29- 00:53 381,358k 7,113k -0.30 d6-d5 e4xd5
29+ 01:47 777,820k 7,205k -0.24 d6-d5 e4xd5 Nd7-b6
30- 01:53 818,806k 7,213k -0.30 d6-d5 e4xd5
30- 02:08 926,543k 7,230k -0.36 d6-d5 e4xd5
30+ 02:33 1,115,401k 7,258k -0.30 d6-d5 a4xb5 c6xb5
31+ 02:56 1,290,340k 7,291k -0.24 d6-d5 e4xd5 Nd7-b6
31- 03:16 1,437,281k 7,311k -0.30 d6-d5 a4xb5
31+ 03:58 1,738,709k 7,305k -0.24 d6-d5 a4xb5 c6xb5
32- 04:15 1,868,829k 7,318k -0.30 d6-d5 a4xb5
32+ 05:30 2,429,403k 7,349k -0.24 d6-d5 e4xd5 Nd7-b6

[bob]

[d]1rbr2k1/2qn1pb1/p1pp2p1/1p2p1Pp/PP1PP3/2P1BP2/3QBNP1/R4RK1 b - - 0 17

In this position, SF played 17...bxa4 ?? and lost the game! After that move, white has a protected passed pawn b4 supported by c3 while black has a very weak backward a6 pawn! I can't understand why bxa4 was played.

Where is this passed pawn white has. With the two black pawns on the a and c files, the white b pawn is not passed. Also black's a6 is not considered "backward" it is considered "isolated".

[Nay Lin Tun]

It was backward in the diagram, and then turned into isolated, "isolated is correct" . . c6 pawn quickly fall after white's d5 move.

[BrendanJNorman]

[d]1rbr2k1/2qn1pb1/p1pp2p1/1p2p1Pp/PP1PP3/2P1BP2/3QBNP1/R4RK1 b - - 0 17

In this position, SF played 17...bxa4 ?? and lost the game! After that move, white has a protected passed pawn b4 supported by c3 while black has a very weak backward a6 pawn! I can't understand why bxa4 was played.

This is the second time this week I seen people pointing out Stockfish's "mistakes" which weren't really mistakes.

17...bxa4 might be (I'm not sure) an inaccuracy, but certainly doesn't deserve "??".

The position is far from static for a start, so generalizing on a position's assessment looking only at pawn structure is a mistake many lower rated players fall into.

There is tension between the d and e pawns and black has the option of trying to liquidate things via ...a5 or ...c5 at some point and then there is the matter of whatever the minor pieces are doing, how strong the bishop on g7 becomes etc.

Remember in chess, a weakness is only a weakness if it can be attacked successfully, so superficial looks at the pawn structure (while many pieces are on the board) will usually lead you astray in your assessment.

Lastly, whites pawn on b4 is protected, but certainly isn't a "passed pawn".

Philidor was wrong. Pawns are not the soul of chess, and thats why 100 years after his prime, the top players were sacrificing them with abandon.

[tpoppins]

SF SSE 161112 in MultiPV=3 mode at high depth considers 17...bxa4 best:

Code: Select all

1.   = (0.27): 17...bxa4 18.dxe5 Nxe5 19.Rxa4 c5 20.Rb1 Qd7 21.Ra2 Qe7 22.Rab2 Be6 23.f4 Nc4 24.Bxc4 Bxc4 25.bxc5 Rxb2 ... 
2. +/= (0.39): 17...d5 18.f4 exd4 19.Bxd4 bxa4 20.f5 Nb6 21.Bxg7 Kxg7 22.fxg6 fxg6 23.Qd4+ Kg8 24.Qf6 Rd6 25.Qe5 Be6 ..
3. +/= (0.53): 17...exd4 18.Bxd4 Ne5 19.Qe3 c5 20.bxc5 dxc5 21.Bxc5 Nc4 22.Bxc4 bxc4 23.Rab1 Rxb1 24.Rxb1 Be5 25.Bb6 Qb8 ...
Depth: 38/77   00:45:04   59563 MN, tb=75536

1.   = (0.23): 17...bxa4 18.Rxa4 d5 19.Qc1 exd4 20.cxd4 dxe4 21.Bf4 Qb7 22.Nxe4 Ra8 23.Bg3 Bxd4+ 24.Bf2 Bxf2+ 25.Rxf2 Nb6 ...
2. +/= (0.40): 17...d5 18.f4 exd4 19.Bxd4 Bxd4 20.cxd4 bxa4 21.f5 Nb6 22.Bd3 Nc4 23.Qe2 Rxb4 24.fxg6 Rb2 25.Qxh5 fxg6 ...
3. +/= (0.50): 17...exd4 18.Bxd4 Ne5 19.Qe3 c5 20.bxc5 dxc5 21.Bxc5 Nc4 22.Bxc4 bxc4 23.Rab1 Rxb1 24.Rxb1 Be5 25.Bb6 Qb8 ...
Depth: 39/77   01:06:23   88030 MN, tb=188176

1.   = (0.21): 17...bxa4 18.dxe5 Nxe5 19.Rxa4 c5 20.Rb1 Qd7 21.Ra3 Qe7 22.Bf4 cxb4 23.cxb4 Bb7 24.Ra5 Nc6 25.Ra4 Nd4 ...
2. +/= (0.63): 17...d5 18.f4 exd4 19.Bxd4 Bxd4 20.cxd4 Nb6 21.axb5 axb5 22.e5 Qe7 23.Qc3 Na4 24.Qb3 Nb6 25.Rfc1 Nc4 ... 
3. +/= (0.65): 17...exd4 18.Bxd4 Ne5 19.Qe3 c5 20.bxc5 dxc5 21.Bxc5 Nc4 22.Bxc4 bxc4 23.Rab1 Rxb1 24.Rxb1 Be5 25.Bb6 Qb8 ...
Depth: 40/77   01:31:23  121496 MN, tb=317602

It is #1 choice from d=30 onward (I ignored the earlier iterations), except for dd=31, 33 and 36 where it gave way to 17...exd4.

On Let's Check there is Komodo 10.2 analysis at d=36 +0.39: 17...d5 18.exd5 Nb6, which is similar to what Jason posted earlier in this thread.

[tpoppins]

Houdini 5.01 Pro:

Code: Select all

1. +/= (0.38): 17...bxa4 18.Rxa4 d5 19.Qc1 exd4 20.cxd4 dxe4 21.fxe4 Nb6 22.Bf4 Qe7 23.Ra5 Rb7 24.Bxa6 Rbd7 25.d5 cxd5 ... 
2. +/= (0.59): 17...exd4 18.Bxd4 Ne5 19.Qe3 bxa4 20.Rxa4 c5 21.bxc5 dxc5 22.Bxc5 Rb2 23.Bd4 Be6 24.Re1 a5 25.f4 Ng4 ...
Depth: 36/79   02:36:24  274051 MN, tb=203712

17...bxa4 doesn't make it to the top of the pile until d=26, then it stays there until the end, except for iterations 29 and 31 (when 17...exd4 temporarily takes over). Its eval has been steadily improving from d=33 on: 0.71, 0.63, 0.44 and finally 0.38.

It's likely that given enough time Komodo will also see 17...bxa4 as best.

[Lyudmil Tsvetkov]

I have to disappoint most of you, as, quite probably, the black position is already lost.

ba4 creates an advanced a4 passer (big plus), but I think this, as well as other possible moves, can not save the game.

and yes, pawns are the soul of chess, it is because of the backward f7 pawn, part of the king shelter, that black is lost; that is a huge penalty, some 50cps or so, but many certainly will not see it.