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

[Carotino]

If there is anyone else who has a rebuttal to 19...Bg4!! to prove Fischer (and me) wrong, I would love to hear from you!, I would like to think that I covered all of the variations, but I know it is very possible that a patzer like me may have missed something that you, my fellow computer chess aficionados may see.

Please, a PGN with all your variations.
Houdini is probably better than Stockfish to analyze such positions (with mating threats).

My PGN, as you can imagine, has been edited so many times over a few months that it is a very sloppy mess and would probably be illegible for most because it is so crappy. I would be embarrassed to post it (so I will not). I prefer that others post a variation they think may refute 19...Bg4!! and then I'll post a rebuttal or I'll congratulate them for busting 19...Bg4!!

19... Bg4!! 20.hxg4 hxg4 21.a6?!

After: 21...Nf3+ 22.Bxf3 gxf3 23.axb7 Qxb7 24.Ng2! fxg2 (24...Bg5!? 25.Ne1=) 25.Kxg2 Kg7 26.f4! is draw.

... I'm sorry for Bobby!

[yanquis1972]

afraid not

[d]r3r1k1/pP1q1p2/2p2npb/1PPp4/3PpN2/2N1PpP1/1R1B1P1K/3Q1R2 b - - 0 23

Analysis by Stockfish 261114 64 BMI2:

23...Bxf4 24.Rh1 Qxb7 25.exf4 Kg7 26.g4 Rh8+ 27.Kg3 Rxh1 28.Qxh1 Rh8 29.Qxh8+ Kxh8 30.b6 Nd7 31.Rb1 Kg7 32.Be3 Kf6 33.Kh4 Ke6 34.Kg3 axb6 35.cxb6 Kd6 36.Na4 Qa8
-/+ (-1.21) Depth: 7/8 00:00:00 4kN
-+ (-2.60) Depth: 25/39 00:00:03 35920kN
23...Ng4+ 24.Kg1 Bxf4 25.Nxe4 dxe4 26.Qxf3 exf3 27.bxa8Q Rxa8 28.e4 Qxd4 29.Bc3 Qxc3 30.e5 Qxb2 31.gxf4 Qd2 32.e6 Qxf4 33.exf7+ Kxf7 34.Re1 Qh2+ 35.Kf1 Qxf2#
-+ (-2.66 ++) Depth: 26/39 00:00:04 47907kN
-+ (-#13) Depth: 35/51 00:00:43 714MN, tb=8

[Carotino]

Wow... You are very speedy!

...Then Bobby was really a genius.

[peter]

22...g5? is no good.

22...Nxg4+ or Bxf4 can transpose but then

22. a6 Ng4+ 23. Kg1 Bxf4 24. exf4 Nh6 25. Kh2 bxa6 26. bxc6 Qxc6 27. Qa2 Ng4+ 28. Kg1 Nf6 29. Kh2 Qd7 30. Qxa6 Qf5 31.Qa2 Kg7 32. Bc1 Rab8 33. h4 a5 34. Rb1 e3 35. fxe3 Qd3

really might be the shortest way and more and more it seems to me almost all of these moves up to 31 are single best moves.

http://www.talkchess.com/forum/viewtopi ... 01&t=54427

The whole thing looks like a study and the engines perform rather poor in getting the tactical shots one after another.
I really wonder how much Fischer might have been able to consider exactly on the board

[acase]

22...g5? is no good.

22...Nxg4+ or Bxf4 can transpose but then

22. a6 Ng4+ 23. Kg1 Bxf4 24. exf4 Nh6 25. Kh2 bxa6 26. bxc6 Qxc6 27. Qa2 Ng4+ 28. Kg1 Nf6 29. Kh2 Qd7 30. Qxa6 Qf5 31.Qa2 Kg7 32. Bc1 Rab8 33. h4 a5 34. Rb1 e3 35. fxe3 Qd3

really might be the shortest way and more and more it seems to me almost all of these moves up to 31 are single best moves.

http://www.talkchess.com/forum/viewtopi ... 01&t=54427

The whole thing looks like a study and the engines perform rather poor in getting the tactical shots one after another.
I really wonder how much Fischer might have been able to consider exactly on the board

Peter, after looking at it I completely agree with your 22.a6 as being the best move to put up the most resistance. It forces white to divert his kingside plans to deal with white's queenside counterplay and keeps the crucial H file closed. You are also correct that it still loses. I agree with almost all of your moves for black with the exception of diverting from your line at move 32.

Let me show you what I mean:

22. a6 Ng4+ 23. Kg1 Bxf4 24. exf4 Nh6 25. Kh2 bxa6 26. bxc6 Qxc6 27. Qa2 Ng4+ 28. Kg1 Nf6 29. Kh2 Qd7 30. Qxa6 Qf5 31.Qa2 Kg7 32. Bc1


[d]r3r3/p4pk1/5np1/2Pp1q1p/3PpP2/2N2bPP/QR3P1K/2B2R2 b - - 0 32

and instead of playing 32...Rab8 here, I would bring the other rook on e8 over by playing 32...Reb8 so I could keep the rook on a8 behind the passed pawn on a7.

[d]rr6/p4pk1/5np1/2Pp1q1p/3PpP2/2N2bPP/QR3P1K/2B2R2 w - - 0 33


Now if 33.h4? is played here we have 33...a5 34.Rb5 e3 and black has a clear victory, so instead of 33.h4 in this position perhaps white should play 33.Rb5 which also loses but takes longer.

[peter]

No program solved it then, and none of the programs I have tried to date have ever solved this position

Just want to tell, that I finally found a SF- setting able to select the best move, even if it takes more than 4 hours with 24 threads and 32 Gb hash.
That's a version of the so called SF mate-finder, Ipman added some adjustable parameters for every single figure, and especially two more seem very interesting, cause to change only the one for middlegames or the other one for endgames to about half up to two thirds of the default value gives better tactical solving in many tricky positions I tried.

r3r1k1/pp1q1p2/2p2npb/PPPp1bnp/3PpN2/2N1P1PP/1R1B1PBK/3Q1R2 b - - 0 1

Analysis by SF Mate Finder+ 64:

19...Tab8 20.a6 cxb5 21.Txb5 bxa6 22.Ta5 Db7
= (-0.10) Tiefe: 7/7 00:00:00 23kN
...
19...Tab8 20.a6 bxa6 21.bxc6 Dxc6 22.Txb8 Txb8 23.Kg1 Dd7 24.h4 Lg4
= (-0.05) Tiefe: 11/15 00:00:00 114kN
...
19...Tec8
= (0.00 ++) Tiefe: 18/24 00:00:06 7795kN
...
19...Tec8 20.a6
= (0.10 --) Tiefe: 22/28 00:00:16 56946kN
19...Sf3+
= (0.06 ++) Tiefe: 22/31 00:00:17 68958kN
...
19...Sf3+ 20.Lxf3
= (0.10 --) Tiefe: 23/33 00:00:19 86587kN
19...a6
= (0.07 ++) Tiefe: 23/33 00:00:19 94417kN
...
19...a6 20.bxc6 bxc6 21.Te1 Dc7 22.Tb6 Tec8 23.Sa4 Te8 24.Sc3
= (0.00) Tiefe: 49/60 02:27:32 162111MN
19...Lg4
= (-0.03 ++) Tiefe: 50/60 04:07:45 275070MN
19...Lg4
= (-0.06 ++) Tiefe: 50/60 04:31:37 302798MN
19...Lg4
= (-0.10 ++) Tiefe: 50/60 04:53:03 327496MN
19...Lg4 20.hxg4
= (-0.06 --) Tiefe: 50/60 05:26:23 365690MN

Sorry for german abbrevations, forgot to change that at beginning,
Of course pv and eval are as well as non-telling, but better than nothing at all.

[acase]

No program solved it then, and none of the programs I have tried to date have ever solved this position

Just want to tell, that I finally found a SF- setting able to select the best move, even if it takes more than 4 hours with 24 threads and 32 Gb hash.
That's a version of the so called SF mate-finder, Ipman added some adjustable parameters for every single figure, and especially two more seem very interesting, cause to change only the one for middlegames or the other one for endgames to about half up to two thirds of the default value gives better tactical solving in many tricky positions I tried.

r3r1k1/pp1q1p2/2p2npb/PPPp1bnp/3PpN2/2N1P1PP/1R1B1PBK/3Q1R2 b - - 0 1

Analysis by SF Mate Finder+ 64:

19...Tab8 20.a6 cxb5 21.Txb5 bxa6 22.Ta5 Db7
= (-0.10) Tiefe: 7/7 00:00:00 23kN
...
19...Tab8 20.a6 bxa6 21.bxc6 Dxc6 22.Txb8 Txb8 23.Kg1 Dd7 24.h4 Lg4
= (-0.05) Tiefe: 11/15 00:00:00 114kN
...
19...Tec8
= (0.00 ++) Tiefe: 18/24 00:00:06 7795kN
...
19...Tec8 20.a6
= (0.10 --) Tiefe: 22/28 00:00:16 56946kN
19...Sf3+
= (0.06 ++) Tiefe: 22/31 00:00:17 68958kN
...
19...Sf3+ 20.Lxf3
= (0.10 --) Tiefe: 23/33 00:00:19 86587kN
19...a6
= (0.07 ++) Tiefe: 23/33 00:00:19 94417kN
...
19...a6 20.bxc6 bxc6 21.Te1 Dc7 22.Tb6 Tec8 23.Sa4 Te8 24.Sc3
= (0.00) Tiefe: 49/60 02:27:32 162111MN
19...Lg4
= (-0.03 ++) Tiefe: 50/60 04:07:45 275070MN
19...Lg4
= (-0.06 ++) Tiefe: 50/60 04:31:37 302798MN
19...Lg4
= (-0.10 ++) Tiefe: 50/60 04:53:03 327496MN
19...Lg4 20.hxg4
= (-0.06 --) Tiefe: 50/60 05:26:23 365690MN

Sorry for german abbrevations, forgot to change that at beginning,
Of course pv and eval are as well as non-telling, but better than nothing at all.

Hi Peter, while it looks like the SF matefinder found the first move here, I don't think it will find a correct winning continuation. The evaluation only shows very slightly better (or worse, I'm not sure what GUI you are using here) than equality in this position which indicates that the program still doesn't understand the position.


If you look earlier in the thread, the poster named "Gary" posted a modified Stockfish version's analysis in which it found the first move also, but was clueless as to how to continue correctly.

By the way, that's a very impressive computer you have there.

[peter]

[ If you look earlier in the thread, the poster named "Gary" posted a modified Stockfish version's analysis in which it found the first move also, but was clueless as to how to continue correctly.

Of course you're right and I well noticed Gary's posting. I will probably find some more settings of different engines with especially for this one and only tuned position "solving" it much faster, if you look for the first move only and no, neither the pv nor the eval say a thing here. But I yet wanted to show this special setting being able without change of the value of a single piece's eval neither nor a special parameter of search like nullmove pruning, to solve several positions I tried of high tactical difficulties

[acase]

But I yet wanted to show this special setting being able without change of the value of a single piece's eval neither nor a special parameter of search like nullmove pruning, to solve several positions I tried of high tactical difficulties

Looks like you succeeded then quite well on this position, and I certainly like the idea.

[Vinvin]

Is there some conclusion for this position now ?

Eval from Stockfish_190726 (3 best moves) after Bg4!? with a bit of learning.

[d]r3r1k1/pp1q1p2/2p2npb/PPPp2np/3PpNb1/2N1P1PP/1R1B1PBK/3Q1R2 w - - 1 20

Code: Select all

Stockfish_19072603_x64_modern:
...
 34/48	01:35	 834 126 572	8 764 135	-1,81	20.Qc1 Nf3+ 21.Bxf3 exf3 22.hxg4 Bxf4 23.exf4 Nxg4+ 24.Kg1 Ne5 25.f5 Nd3 26.Qb1 Qxf5 27.Kh2 Nxb2 28.Qxb2 h4 29.Qc1 Qg4 30.Bg5 hxg3+ 31.fxg3 f6 32.Bxf6 Kf7 33.Qf4 Qxf4 34.gxf4 Re3 35.Nxd5 cxd5 36.Be5 Rb3 37.Kg3 Rxb5 38.Kxf3 Rxa5 39.Rh1 Ke6 40.Rh7 Rb5 41.Rg7 Rb3+ 42.Kg4
 34/51	01:35	 834 126 572	8 764 135	-1,33	20.hxg4 hxg4 21.Rh1 Nf3+ 22.Bxf3 gxf3 23.Nh3 Kg7 24.Qf1 Rh8 25.Kg1 g5 26.Ra2 g4 27.a6 gxh3 28.Qxh3 Qxh3 29.Rxh3 Bxe3 30.Rh4 Bxd2 31.Rxd2 Rhb8 32.axb7 Rxb7 33.bxc6 Rc7 34.Ra2 a5 35.Kf1 Rxc6 36.Ke1 Kg6 37.Na4 Re6 38.Nb6 Rae8 39.Rxa5 e3 40.Ra7 exf2+ 41.Kxf2
 35/54	01:35	 834 126 572	8 764 135	-0,85	20.Qb1 Nf3+ 21.Bxf3 Bxf3 22.Ng2 a6 23.bxc6 bxc6 24.Rb7 Qe6 25.Ne1 Ng4+ 26.hxg4 hxg4 27.Rh1 Kg7 28.Kg1 Bxh1 29.Kxh1 Bg5 30.Kg1 Rh8 31.Nd1 Qf5 32.Bc3 Rh3 33.Rb8 Rxb8 34.Qxb8 Bf6 35.Kf1 Qh5 36.Ke2 Rh1 37.Qb7 Qh3 38.Kd2 Qf1 39.Kc1 Qc4