Test Position: Nikolic Fischer revisited 19...Bg4!!

[acase]

This position was posted a little more than 6 years ago (when Depp Rybka 3 reigned supreme) and there was a difference of opinion as to whether the move actually won or not. The late late Robin Smith who was a correspondence GM was of the opinion that the move probably won, which I am in complete agreement with after researching the position for quite a long time.


of course I tried several different replies for white including 20.Qc1, 20.Qa1, 20.Qb1, 20.Qc2, 20.Qb3,20.Qa4, 20.Nxg6, 20.Nfe2, 20.Nce2, 20.Qe1, and 20.f3. I have come to the conclusion that all replies by white lose no matter what. The gentleman at the end of the 6 year old thread named "Mischa" had it all correct as to what Fischer's winning plan was, also it must be pointed out that Fischer himself says the 19.bg4!! wins.

No program solved it then, and none of the programs I have tried to date have ever solved this position (including all of the new ones e.g. Komodo8, Latest Stockfish, Critter, Houdini 4, and Gull 8).

Here is the original thread:
http://talkchess.com/forum/viewtopic.ph ... ic+fischer


And here is the position in question:

[d]r3r1k1/pp1q1p2/2p2npb/PPPp1bnp/3PpN2/2N1P1PP/1R1B1PBK/3Q1R2 b - - 0 19

[rabbits23]

Andrew: Thanks for the position; it will give me something to think about
over the next few days. It is certainly fairly complicated and I would have
said fairly drawish but clearly Fischer would disagree. I read some of the
comments on the original web page and I thought there were a number of clever, thoughtful observations. I particularly liked chrisw's comment when he said, "Chess can be like picking a path through a minefield. One wrong turn and bang." But then again Fischer was an experienced, hardened campaigner and might have thrived on positions like this. My
Stockfish 2.3 program does not find 19... Bg4 and when I enter the
variation !9]... Bg4 20) hxg4 hxg4 21) Rh1 Nf3+ 22)Bx f3 gxf3 it has
white ahead +- (2.50) Still it's early days yet and I'll see how things pan
out with closer analysis.
Allan

[rabbits23]

Sorry about the text muck up in the first post. I forgot to preview it
Allan

[gladius]

I didn't even try normal SF on this, but a version with some king attack evaluation meant to help with kings indian type positions does find Bg4. Not with a winning score, but it still picks the move. Now to get this code into mainline SF .

...
info depth 31 seldepth 43 multipv 1 score cp 9 lowerbound nodes 728577747 nps 3891164 time 187239 pv f5g4
info depth 31 seldepth 43 multipv 1 score cp 30 lowerbound nodes 999814056 nps 3855031 time 259353 pv f5g4
info depth 31 seldepth 48 multipv 1 score cp 14 upperbound nodes 1544103788 nps 3820715 time 404140 pv f5g4 d1b1
info depth 31 seldepth 49 multipv 1 score cp 14 nodes 1764781865 nps 3806682 time 463601 pv f5g4 d1b1 g4f3 a5a6 f3g2 h2g2 g5f3 a6b7 d7b7 b1c2 h6f4 g3f4 e8c8 f1b1 b7d7 b5c6 c8c6 b2b7 d7c8 c3d5 f6d5 c2e4 f3h4 g2h2 c8e6 e4e6 c6e6 d2a5 h4f5 b7b8 a8b8 b1b8 g8g7 b8d8 d5f6 d4d5 e6a6 a5c3 a6a2 e3e4 a2f2 h2g1 f2c2 c3f6 g7f6 e4f5

[Lyudmil Tsvetkov]

This position was posted a little more than 6 years ago (when Depp Rybka 3 reigned supreme) and there was a difference of opinion as to whether the move actually won or not. The late late Robin Smith who was a correspondence GM was of the opinion that the move probably won, which I am in complete agreement with after researching the position for quite a long time.


of course I tried several different replies for white including 20.Qc1, 20.Qa1, 20.Qb1, 20.Qc2, 20.Qb3,20.Qa4, 20.Nxg6, 20.Nfe2, 20.Nce2, 20.Qe1, and 20.f3. I have come to the conclusion that all replies by white lose no matter what. The gentleman at the end of the 6 year old thread named "Mischa" had it all correct as to what Fischer's winning plan was, also it must be pointed out that Fischer himself says the 19.bg4!! wins.

No program solved it then, and none of the programs I have tried to date have ever solved this position (including all of the new ones e.g. Komodo8, Latest Stockfish, Critter, Houdini 4, and Gull 8).

Here is the original thread:
http://talkchess.com/forum/viewtopic.ph ... ic+fischer


And here is the position in question:

[d]r3r1k1/pp1q1p2/2p2npb/PPPp1bnp/3PpN2/2N1P1PP/1R1B1PBK/3Q1R2 b - - 0 19

Very interesting position, Andrew.

Sorry to post without substantiating my claims by any output, but intuitively I say black is winning. Bg4 is the first move I would think of.

An open h file is twice as dangerous as other open files. I would assess black advantage at some 70cps, which is huge in this rich middlegame. What is losing white is the b7-c6-d5-e4 KID chain, and the fact that the f2 pawn, part of the king shelter, is very weak, backward. The weak f3 square is the key to the black success.

Only person that could tell you straight away is black is winning is Alekhine: he was much much better than Kasparov in such positions, not to mention Fischer. Fischer, of course, knew his business.

[carldaman]

I didn't even try normal SF on this, but a version with some king attack evaluation meant to help with kings indian type positions does find Bg4. Not with a winning score, but it still picks the move. Now to get this code into mainline SF .

Now that's promising, Gary

Cheers,
CL

[Lyudmil Tsvetkov]

It proves this is altogether a mate in 5 or 6; at least black needs only 5 or 6 moves to reach a full 3 pawns material advantage.

SF on my computer does not find Bg4, but when the move is played, it finds an easy win both after hg4 hg4, and after Qb1 Nf3. The win is found in a shootout, not in analysis, with SF giving at first uncertain scores in some lines, but still winning as the game progresses.

Btw., in Gary's modified SF PV, after Qb1 Nf3, not Bf3 is the right move, followed by Bf3 Bf3 and Ng4 or h4 for black (or exchanging on f4 before that)

Again, black has huge advantage here.

[acase]

Sorry about the text muck up in the first post. I forgot to preview it
Allan

No problem, I think we all knew what you meant. I mucked my original post up a little bit...as there is no such thing as "Gull 8", and when I stated that "Fischer himself says the 19.Bg4!! wins (should have been "Fischer himself says that 19...Bg4!! wins)

[acase]

It proves this is altogether a mate in 5 or 6; at least black needs only 5 or 6 moves to reach a full 3 pawns material advantage.

SF on my computer does not find Bg4, but when the move is played, it finds an easy win both after hg4 hg4, and after Qb1 Nf3. The win is found in a shootout, not in analysis, with SF giving at first uncertain scores in some lines, but still winning as the game progresses.

Btw., in Gary's modified SF PV, after Qb1 Nf3, not Bf3 is the right move, followed by Bf3 Bf3 and Ng4 or h4 for black (or exchanging on f4 before that)

Again, black has huge advantage here.

Hello Lyudmil, Yes after 20.hxg4 hxg4 which was played in the original game white loses even if he tries 21.b5xc6 attacking the black queen and trying for queenside counter play, 21.Nxg6 trying for counter play and giving the piece back, or as was tried in the game 21.Rh1.

You are also correct about Gary's modified SF PV.

[rabbits23]

Andrew: I think you are right and that Black does win if White accepts the
Bishop sacrifice . I wasn't aware of Fischer's plan until I read Mischa
post in the previous web page. I played it out against SF which does not
recognise that it is losing till move 28. Here is how the game proceeded
(19)... Bg4 (20) h3xg4 h5xg4 (21) Rh1 Nf3+ (22) BxN g4xf3 (23)Nh3
...Kg7 (24) Kg1 Rh8 (25) Qf1 Rh7 (26) a6 R(a8)h8 (27)axb7 Qg4 (28)b6
Bf4 (- 3.45)
It might be that some other computer chess program can find a way of outsmarting
Fischer's plan but I would doubt it.
Regards Allan