Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Very hard position for Rybka 3 (Fischer again!)
- - By sylvain.renard.27 (**) [fr] Date 2008-09-07 15:49
r3r1k1/pp1q1p2/2p2npb/PPPp4/3Pp3/2NNPpP1/1R1B1P1K/3Q3R b - -


During the Game Nikolic-Fischer (Vinkovci, 1968). White played 23.Kg1. Instead of 23.Kg1, if you play 23.Nf4d3, 23... exd3 leads to a draw. The winning move is 23...Kg7! (Fischer)

Rybka 3 x64, multiprocessor Q6600 4x3600 MHz, 1024 Mb of hash tables.

..........
23...exd3 24.Qxf3 Ne4 25.Kg2 Bg5 26.a6 cxb5 27.axb7 Rab8 28.Rxb5 Nxc3 29.c6 Qxc6 30.Rc5[] Qxb7 31.Rxc3 Kg7 32.Rhc1 Rec8 33.Rxc8 Rxc8 34.Rxc8 Qxc8 35.Qxd5[] Qc2 36.Qxg5 Qxd2 37.Qe5+ Kh7 38.d5 Qa2
  =  (0.00)   Depth: 24   00:58:05  747mN
23...exd3 24.Qxf3 Ne4 25.Kg2 Bg5 26.a6 cxb5 27.axb7 Rab8 28.Rxb5 Nxc3 29.c6 Qxc6 30.Rc5[] Qxb7 31.Rxc3 Kg7 32.Rhc1 Rec8 33.Rxc8 Rxc8 34.Rxc8 Qxc8 35.Qxd5[] Qc2 36.Qxg5 Qxd2[] 37.Qe5+ Kh7 38.d5 Qa2
  =  (0.00)   Depth: 25   02:08:30  1735mN
23...exd3 24.Qxf3 Kg7 25.a6 b6 26.Kg2 Bg5
  =  (0.00)   Depth: 26   10:52:02  9512mN, tb=41
23...Kg7 24.bxc6 bxc6[]
  =  (-0.15)   Depth: 26   19:12:33  17756mN, tb=42
Parent - - By sylvain.renard.27 (**) [fr] Date 2008-09-08 17:10
23...Kg7
  =/+  (-0.35 !)   Depth: 27   40:03:28  40594mN, tb=181
Parent - - By sylvain.renard.27 (**) [fr] Date 2008-09-09 10:29
23...Kg7
  =/+  (-0.55 !)   Depth: 27   59:40:05  62653mN, tb=182
Parent - - By sylvain.renard.27 (**) [fr] Date 2008-09-10 19:57
23...Kg7
  -/+  (-0.95 !)   Depth: 27   95:54:52  102701mN, tb=185
Parent - - By Regicide (*) [us] Date 2008-09-10 22:35

> 23...Kg7
>   -/+  (-0.95 !)   Depth: 27   95:54:52  102701mN, tb=185


I think it's a good idea to see how long it takes Rybka to see the win. But could you post the line(s) it sees also?
Parent - - By sylvain.renard.27 (**) [fr] Date 2008-09-14 17:00
23...Kg7
  -+  (-1.75 !)   Depth: 27   186:45:56  204379mN, tb=226
Parent - By nine castles (**) [us] Date 2008-09-17 12:00
That is very interesting. I don't think I've seen a position where an engine evaluates the position as winning after so long: I think most testers would give up by that point :)
Parent - By Vinvin (***) [be] Date 2008-09-17 12:20
Still in depth 27, still only 1 move but eval more and more convincing !
Parent - By sylvain.renard.27 (**) [fr] Date 2008-09-14 17:14
Sorry, there is no line yet! It seems Rybka 3 uses "aspiration search" At depth 26, 23...Kg7 +0.15. At depth 27, the search is between +0.15 -0.20 = -.05 and +.15 + 0.20 = +0.35
She finds the move 23...Kg7 is better than +0.35; so she searches again between  +0.35 -0.20 = 0.15 and +0.35 + 0.20 = +0.55 . It was better than +.055, another search between +0.55 -0.40 = +0.15 and +0.55 + 0.40 = +0.95. Next search  was between +0.95 - 0.80 = 0.15 and +0.95 +0.80 = +1.75. Now she is searching between +1.75 - 1.60 = +0.15 and +1.75 + 1.60 = +3.35 but it is not finished and there will be no line if the move 23...Kg7 is better than the bound.
  Best regards,
    Sylvain
Parent - - By srwike (*) [us] Date 2008-09-10 23:35
Did you try Rybka 3 Dynamic using default settings? In one case (http://www.chessville.com/misc/whychessengines.htm midway down the page) where this author claimed the top engines would not find the win but the old ones would, I tested all three Rybka 3 engines. Rybka 3 standard could not find it in 13 minutes (I gave up) but dynamic found it in 20 seconds and human in 45 seconds.

Just a thought.
Parent - By sylvain.renard.27 (**) [fr] Date 2008-09-14 17:15
No, it is standard Rybka 3.
Parent - - By ptobler (**) [au] Date 2008-09-17 12:39
My copy of "Bobby Fischer Rediscovered" has:

"23...Bxf4! (Fischer's move)

Not 23...Kg7 because 24 Rxh6! Kxh6 25 a6 gives White chances he doesn't deserve"
Parent - - By sylvain.renard.27 (**) [fr] Date 2008-09-17 18:00
Hello,
we are not talking of the same position, because after 23...Kg7, white cannot play Rxh6 (White king is on h2).
  Best regards,
     Sylvain
 
Parent - - By Jim Logan (*) [us] Date 2008-09-17 22:10
I see no position as you say - here is the game you reference:

[Event "Vinkovci"]
[Site "Vinkovci"]
[Date "1968.??.??"]
[Round "9"]
[White "Nikolic, Emil"]
[Black "Fischer, Robert James"]
[Result "0-1"]
[ECO "E61"]
[PlyCount "62"]
[EventDate "1968.09.??"]
[EventType "tourn"]
[EventRounds "13"]
[EventCountry "YUG"]
[Source ""]
[SourceDate ""]

1. c4 g6 2. Nc3 Bg7 3. g3 e5 4. Bg2 d6 5. e3 Nf6 6. Nge2 O-O 7. O-O c6 8. d4
Re8 9. Rb1 e4 10. b4 Bf5 11. h3 h5 12. Nf4 Nbd7 13. a4 Nf8 14. c5 d5 15. b5
N8h7 16. Bd2 Ng5 17. Rb2 Qd7 18. Kh2 Bh6 19. a5 Bg4 20. hxg4 hxg4 21. Rh1 Nf3+
22. Bxf3 gxf3 23. Kg1 Bxf4 24. exf4 Kg7 25. f5 Rh8 26. Bh6+ Rxh6 27. Rxh6 Kxh6
28. Qd2+ g5 29. bxc6 Qxf5 30. Nd1 Qh3 31. Ne3 Kg6 0-1
Parent - By oudheusa (*****) [nl] Date 2008-09-18 07:21
I was confused at first too, but apparently this is analysis for the position after 23.Nd3 instead of 23.Kg1.
Parent - By oudheusa (*****) [nl] Date 2008-09-17 18:42 Edited 2008-09-17 18:58
what is Fischer's or Rybka's winning line after 23.. Kg7 ?

I can't find anything after 24. Ne5 Ng4+ 25. Kg1!
Parent - - By oudheusa (*****) [nl] Date 2008-09-18 07:23
Hello Sylvain, can you plse give Fischer's line after Kg7. Which source did you find this in?
Parent - - By sylvain.renard.27 (**) [fr] Date 2008-09-18 12:24
Hello,
some people thought that 23.Nf4d3 was the refutation of Fischer's attack, because they did not let Rybka enough time on this position! They only saw fxd3? which leads to a draw. But the move 23...Kg7! gives black winning chances. It seems that nobody saw that 23...Kg7! was the good move. Really nobody? As Fischer played this game, I supposed that he would have played it, but of course, I have no proof!
  Rybka 3 is very very long to display the line at depth 27.  Previous evaluations were given after 96 hours, and 187 hours, without any line because the evaluation was greater than the bound (the last one is 1.75 for black). I hope that the next one will be smaller than 3.35 for black, in this case Rybka 3 will give us a line. But it will take more than 300 hours anyway.
  Best regards,
     Sylvain 
Parent - - By oudheusa (*****) [nl] Date 2008-09-18 12:28
ah ok, it seemed to me you suggested Fischer had analysed Kg7 and thought it was winning. I look forward to some lines, because I think white is still ok and will get at least a draw.
Parent - - By sylvain.renard.27 (**) [fr] Date 2008-09-18 13:49
Rybka sees more or less +2 for black after 23...Kg7. So I think it is at least a draw for black!
Parent - - By sylvain.renard.27 (**) [fr] Date 2008-09-19 20:52
As the line was very very long to come, I played 23...Kg7 and 24.bxc6. Here is the line found by Rybka 3 :
24...Ng4+ 25.Kg1[] bxc6[] 26.Ne5 Nxe5[] 27.dxe5 Bg5[] 28.Ne2 Rh8 29.Rxh8 Rxh8 30.e6[] fxe2[] 31.Bc3+ Bf6 32.Bxf6+ Kxf6 33.Qxe2 Qxe6
  -+  (-1.94)   Depth: 23   06:59:09  7724mN, tb=98
Parent - By nine castles (**) [us] Date 2008-09-20 06:52
It may be worth it to run monte carlo on the position with very high depth, since you are willing to let the machine think for days anyway, this may produce more definitive results.
Parent - - By oudheusa (*****) [nl] Date 2008-09-20 08:01
28.Ne2? is a very strange move; my Rybka 3 does not play it.

After 28. Na4! Rh8 29. Rxh8 Rxh8 30. e6 white is better:

Rybka 3

[+0.97]  d=14  8...Qxe6 9.Bc3 f6 10.Rb7 Kg8 11.Rb8 Kg7 12.Rxh8 Kxh8 13.Kh2 Kg7 14.Qf1 Qd7 15.Qh3 Qb7 16.Bd4 Kg8 17.Qe6 Kg7 18.Qg4 (0:00.57)
[+1.18]  d=13  8...Qc8 9.Bc3 Bf6 10.Bxf6 Kxf6 11.Qd4 Kxe6 12.Nc3 f6 13.Nxe4 dxe4 14.Qxe4 Kf7 15.Qxf3 Rh5 16.e4 Rxc5 17.Qb3 Qe6 18.Qb7 Qe7 19.Qxe7 Kxe7 20.Rb7 Kd6 21.Rxa7 f5 22.exf5 gxf5 23.Kg2 (0:00.49)
[+1.58]  d=13  8...Qe7 9.Bc3 f6 10.Qb1 Qxe6 11.Rb7 Kh6 12.Qf1 Qg4 13.Kh2 Rh7 14.Rxh7 Kxh7 15.Qh3 Qxh3 16.Kxh3 f5 17.Nb6 Be7 18.Nd7 (0:00.51)
Parent - - By sylvain.renard.27 (**) [fr] Date 2008-09-20 08:58 Edited 2008-09-20 09:51

> After 28. Na4! Rh8 29. Rxh8 Rxh8 30. e6 white is better:


That's why Rybka 3 plays 28...Qe6! and sees Black winning after this move (3.20 for black at the moment, I'll post the line when it comes)
Parent - By oudheusa (*****) [nl] Date 2008-09-20 09:43
I see, yes 28.. Qe6 is much stronger. It seems black is winning indeed. Very interesting and complicated position!
Parent - By oudheusa (*****) [nl] Date 2008-09-20 08:44
I even found a study like win in the variation above; from the diagram position:

1... Kg7 2. bxc6 Ng4+ 3. Kg1 bxc6 4. Ne5 Nxe5 5. dxe5 Bg5 6. Na4 Rh8 7. Rxh8
Rxh8 8. e6 Qxe6 9. Bc3+ f6 10. Rb7+ Kg8 11. Rb8+ Kg7 12. Rxh8 Kxh8 13. Qf1 Kg8
14. Nb6 {!!} +-

The knight can not be captured and after eg f5 or Bh6 follows Qa6!
Parent - By CSullivan (**) [us] Date 2008-11-13 16:20
Hi Sylvain,
I've really enjoyed this thread about the 23rd move in Nikolic-Fischer.  Thank you for the great analysis of 23.Nd3 Kg7! etc.  I have Rybka (2.3.2a mp 64-bit on an Intel Q6600@3GHz) analyzing a different move: 23.Ng2
r3r1k1/pp1q1p2/2p2npb/PPPp4/3Pp3/2N1PpP1/1R1B1PNK/3Q3R b - -

So far, Rybka gives ....
(-0.54)  Depth:28  06:16:14  10648mN  23...Bg5 24.Qf1 Ng4+ 25.Kg1 fxg2 26.Qxg2 Kg7 27.a6 bxa6 28.bxa6 Qxc6 29.f4 exf3 30.Qxf3
Perhaps you are interested in also analyzing this move?
Regards,
Charles Sullivan
Parent - - By fejman (**) [pl] Date 2008-11-13 17:10
df10  ..Kg7  after 4 min !! on my old dual core
Parent - By CSullivan (**) [us] Date 2008-11-18 17:10
Indeed, the plan to attack on the h-file works just as well after 23.Ng2 as it does after 23.Nh3 or 23.Nd3.  Play might go 23.Ng2 Kg7 24.Nh4 Rh8 25.Qf1 g5 26.Nxf3 exf3 27.Kg1 g4
r6r/pp1q1pk1/2p2n1b/PPPp4/3P2p1/2N1PpP1/1R1B1P2/5QKR w - - 2 28

and White is quite lost.  The concluding moves might be 28.Qc1 Qf5 29.Rh4 Bg5 30.e4 Nxe4 31.Bxg5 Nxg5 32.a6 Nh3+ 33.Kh2 Rae8 34.axb7 Rxh4 35.gxh4 Qf6 36.b8=Q Rxb8 37.h5 Rh8
7r/p4pk1/2p2q2/1PPp3P/3P2p1/2N2p1n/1R3P1K/2Q5 w - - 2 38

and White, just one pawn down, has a hopeless position.
Up Topic Rybka Support & Discussion / Rybka Discussion / Very hard position for Rybka 3 (Fischer again!)

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill