Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka can't find 24.Rxd4!
- - By Zherkovv (***) Date 2008-11-13 10:23
From the game: Kasparov - Topalov
b2r3r/k4p1p/p2q1np1/NppP4/3p1Q2/P4PPB/1PP4P/1K1RR3 w - - 0 24


347: Kasparov,G - Topalov,V, Hoogovens Wijk aan Zee 1999
b2r3r/k4p1p/p2q1np1/NppP4/3p1Q2/P4PPB/1PP4P/1K1RR3 w - - 0 1


Analysis by Rybka 3 32-bit:

24.Nc6+ Bxc6[] 25.Qxd6 Rxd6[] 26.dxc6 Rxc6[] 27.Re7+ Kb6[] 28.Rxf7 Re8 29.Kc1 Rd6 30.f4 c4 31.Bg2 Ng4 32.b4 cxb3 33.cxb3 Ne3 34.Rb7+ Ka5 35.Rd2 Rc8+[] 36.Kb2 Nxg2[] 37.Rxg2 d3[] 38.Rd2 h5[] 39.a4
  =/+  (-0.32)   Depth: 16   00:00:00  0kN
24.Nc6+ Bxc6[] 25.Qxd6 Rxd6[] 26.dxc6 Rxc6[] 27.Re7+ Kb6[] 28.Rxf7 Re8 29.Kc1 Rd6 30.f4 c4 31.Bg2 Ng4 32.b4 cxb3 33.cxb3 Ne3 34.Rb7+ Ka5 35.Rd2 Rc8+[] 36.Kb2 Nxg2[] 37.Rxg2 d3[] 38.Rd2 h5[] 39.a4
  =/+  (-0.32)   Depth: 17   00:00:00  0kN

(,  13.11.2008)

Deep Fritz 11 finds 24.Rxd4 in less than 60 sec.!
Parent - - By Permanent Brain (*****) Date 2008-11-13 10:47
I had Spike 1.2 loaded at the moment, so I tried it. It smells it at depth 16, but unfortunately skips it.

Analysis by Spike 1.2 Turin (D945 3.4 GHz 256 MB hash):

24.Qxd6 Rxd6
  +/-  (1.12)   Depth: 1   00:00:00
(...)
24.Nc6+ Bxc6 25.Qxd6 Rxd6 26.dxc6 Rxc6 27.Re7+ Ka8 28.Rxf7 Re8 29.b4
  =  (-0.03)   Depth: 8   00:00:00  218kN
(...)
24.Nc6+ Bxc6 25.Qxd6 Rxd6 26.dxc6 Rxc6 27.Re7+ Kb6 28.Rxf7 Re8 29.Rd2 Rd6 30.Ka2 c4 31.b4 cxb3+ 32.Kxb3 Re3+ 33.Ka2 Rxf3 34.Re2
  =/+  (-0.63)   Depth: 16   00:04:53  190mN

I didn't wait for depth 17 but I'd guess Spike will decide for Rxd4 then. My cpu is rather slow.

Rybka WinFinder 2.2 is fast here:

Analysis by Rybka WinFinder 2.2 32-bit:

24.Qxd6
  =  (0.00)   Depth: 4   00:00:00
24.Qxd6
  =/+  (-0.41)   Depth: 5   00:00:00
24.Nc6+ Bxc6 25.dxc6
  =  (-0.06)   Depth: 5   00:00:00
(...)
24.Rxd4
  =  (-0.13)   Depth: 10   00:00:05  234kN
24.Rxd4 cxd4 25.Re7+ Kb6 26.Qxd4+ Kxa5 27.Qc3+ b4 28.axb4+ Qxb4 29.Qc7+ Ka4 30.b3+ Ka3
  =  (-0.13)   Depth: 10   00:00:06  411kN
(...)
24.Rxd4 Kb6 25.b4 Qxf4 26.Rxf4 Nxd5 27.Rxf7 cxb4 28.axb4 Nxb4 29.Nb3 Bd5 30.Rf6+ Nc6
  =  (-0.13)   Depth: 15   00:01:55  7459kN
Parent - - By benben (***) [au] Date 2008-11-13 11:04
How do we know Rxd4 really is the best move?
Parent - - By M ANSARI (*****) [kw] Date 2008-11-13 11:19
Exactly ... in this game I remember that there was a drawing line for Topalov.  But instead Topalov decided to call Kasparov's bluff and ended up losing spectacularly.
Parent - By Vempele (Silver) [fi] Date 2008-11-13 11:32
Not drawing but equalizing. IIRC black has the advantage after any other move.
Parent - By Debaser (***) Date 2008-11-13 11:42
Kasparov comments on this move are here

24.Rxd4!!

When I made this move, I saw only the repetition of the moves and the opportunity to continue the attack, though the whole picture of the combination was not yet clear. I already saw the idea 30…Rd6 31.Rb6, but I still could not get rid of the thought that all lines should be checked to the very end. Maybe Black will find some opportunity for defense. Topalov spent about 15 minutes thinking.

I walked around the hall – rather, I fled – and at these feverish moments it seemed to me that there were very few participants and that most of the games had already been finished. My mind worked only in one direction, and one of these moments brought me the image of the whole cluster of various lines. I saw the move 37.Rd7. I don't even remember how this line was formed in my head, but I saw the whole line up to the end. I saw the journey of the black King after 36.Bf1, 37.Rd7 and I could no longer suppress my excitement, because at that same moment I realized that the move 24…Kb6 ruined the whole construction.

The mere thought that I could spoil such a combination drove me crazy, and I only prayed that Topalov would capture on d4. I still was not sure that this would win, but the beauty of the combination I saw impressed me.

I could not believe my own eyes when Veselin twitched abruptly and grabbed the Rook. As he explained after the game, he was exhausted by the tense fight and he thought that White would have to force a draw by the repetition of moves after the Rook was captured. He saw the main idea of the combination, but it did not occur to him that White would play without the Rook, trying to make use of the King's forward position on a4.

24…cxd4?!

This move loses the game, but it is worth an exclamation mark, as great combinations cannot be created without partners. If Topalov had not taken the Rook, the game could have finished in a draw: Veselin would have had half a point more, I – half a point less. He would have win a little bit, I would have lost a little bit, but chess and chess amateurs would have lost a lot. However, Caissa was kind to me that day… I do not know what I was rewarded for, but the development of events became forced after the capture on d4.
Parent - By stvs (***) [gr] Date 2008-11-13 11:31
i think Rxd4 is good move but not as winning move..so the remark "!!" should be "!?" thats why rybka prefer other move..topalov pxd4 is mistake.
also its not the game of century from my opinion...try this with rybka :)

[Event "Linares"]
[Site "?"]
[Date "1999.??.??"]
[Round "?"]
[White "Topalov, V."]
[Black "Ivanchuk, V."]
[Result "0-1"]
[ECO "A30"]
[WhiteElo "2700"]
[BlackElo "2710"]
[PlyCount "50"]
[EventDate "1999.??.??"]

1. Nf3 c5 2. c4 Nc6 3. d4 cxd4 4. Nxd4 e6 5. g3 Bb4+ 6. Nc3 Qa5 7. Ndb5 d5 8.
a3 Bxc3+ 9. bxc3 Nf6 10. Bg2 O-O 11. Qb3 dxc4 12. Qxc4 e5! 13. Nd6 Be6 14. Qd3
e4!! 15. Nxe4 Nxe4 16. Bxe4 Rad8 17. Qc2 Nd4!! 18. Qb2 Nxe2!! 19. Kxe2 Rfe8!!! 20. Qb4
Qh5+! 21. f3 f5! 22. g4 Qh3! 23. gxf5 Bxf5 24. Qc4+ Kh8 25. Re1 Rxe4+! 0-1

rybka found all the great moves here...

this is the game of century of my opinion
Parent - - By Arsha (**) Date 2008-11-13 12:27
As I commented in another thread, Rxd4 is only good for a human opponent (in which black (the human opponent) accepts the sac). However many engines will choose Rxd4 at some reasonable time control. Here is a test have done by Mr. Corbit.
Parent - - By Kappatoo (*****) [de] Date 2008-11-13 12:41
Is there another move for white which equalizes?
Parent - - By Arsha (**) Date 2008-11-13 12:48 Edited 2008-11-13 12:51
Note: accepting the sacrifice is mistake. Leading to drawing situation.
Parent - - By Kappatoo (*****) [de] Date 2008-11-13 14:56
Sorry, I'm not sure what you mean. Doesn't accepting the sac lose for black?
What I tried to say was this: Doesn't Rxd4 objectively lead to an equal position (after black's Kb6) and at the same time, isn't it true that white's alternatives all lead to positions which are worse for white?
Parent - - By Arsha (**) Date 2008-11-13 15:51

> Sorry, I'm not sure what you mean. Doesn't accepting the sac lose for black?


Doesn't accepting the sac give equal chances for both (with a little favor for black) but accepting the sac is draw. Here is the logical finish by Cyclone 2.2 at fixed depth of 14 and after accepting the sac:

25. Re7+ Kb6
26. Qxd4+ Kxa5
27. b4+ Ka4
28. Qc3 Qxd5
29. Qc7 Qd1+
30. Kb2 Qd4+
31. Kb1 Qg1+
32. Kb2 Qd4+
33. Kb1 Qg1+
34. Kb2 Qd4+ (draw)

> isn't it true that white's alternatives all lead to positions which are worse for white?


I don't think so (Nobody is sure about it). In an eng-eng match I think 24.Nc6+ is better than 24.Rxd4. Many engines including Rybka prefer 24.Nc6+. However I think both alternatives give equal chances for black and white (Or results are a little biased in favor of black). It could be tested by an eng-eng match simulation, but I'm not going to examine this. With a human opponent I prefer 24.Rxd4. :)
Parent - - By Vempele (Silver) [fi] Date 2008-11-13 15:55 Edited 2008-11-13 16:05
29. Ra7 Bb7 30.Rxb7 Rhe8 31.Rb6 Ra8 32.Bf1

CAP isn't minimaxed?? The score jumps from 0.00 (Ra8) to 1.33 (Bf1).
Parent - - By Arsha (**) Date 2008-11-13 18:01
Which engine did you use for analysis? I am using slow hardware but I think deeper analysis could shed some facts.
Parent - By Vempele (Silver) [fi] Date 2008-11-13 18:04
WinFinder.

The best line nets White three pawns for the Exchange.
Parent - - By Kappatoo (*****) [de] Date 2008-11-13 17:54
Sorry, but fixed depth 14 is totally insufficient to examine this position. I think it was proven back then with extremely extensive analysis that white wins after cxd4. (Okay, I'm not sure here. I think there was one line where black gets a very bad, but not definitely lost endgame.)

Edit: And the line you gave above indicates that Nc6+ is evaluated as -0.32 by Rybka.
Parent - By Arsha (**) Date 2008-11-13 18:25
I have had some deeper analysis with other engines and after cxd4 position, the results indicated a draw. With this case, I used Cyclone just for showing the draw. Ok, I accept that extremely extensive analysis showed the white win after cxd4. As I said before, accepting the sacrifice is suboptimal.

> And the line you gave above indicates that Nc6+ is evaluated as -0.32 by Rybka.


What evaluation did Rybka give for Rxd4?
(IIRC with 2 line multi-pv and depth 20, R3 preferred NC6+.)
Parent - - By Pia (****) [ru] Date 2008-11-13 12:16

>I didn't wait for depth 17 but I'd guess Spike will decide for Rxd4 then. My cpu is rather slow.


No, Spike ain't found it on even 20th depth...
Parent - By peter2804 (**) [gb] Date 2008-11-14 01:59
I used Junior 7 (my only Junior) for the position but it didn't look at Rxd4 until i changed the parameters to 'favour sacrifices' and it found Rxd4 after 2 secs. It stayed with the move and played it after 3 mins  (time 30 min game) i then removed the ' sacrifice parameter' and after looking at PxR it changed to and played Bxd5 for black followed by whites QxQ  = Do stronger Juniors have the 'sacrifice parameter'?
Parent - - By Zherkovv (***) Date 2008-11-14 07:26
Thanks!
Parent - By StrongHold (*) [us] Date 2008-11-14 19:40
New game - Rybka 3
b2r3r/k4p1p/p2q1np1/NppP4/3p1Q2/P4PPB/1PP4P/1K1RR3 w - - 0 1


Analysis by Rybka 3 :

24.Nc6+ Bxc6 25.Qxd6 Rxd6 26.dxc6 Rxc6 27.Re7+ Kb6 28.Rxf7 Re8 29.Kc1 Rd6 30.f4 c4 31.Bg2 Re2
  =  (-0.23)   Depth: 13   00:00:00  1kN
24.Nc6+ Bxc6 25.Qxd6 Rxd6 26.dxc6 Rxc6 27.Re7+ Kb6 28.Rxf7 Re8 29.Kc1 Rd6 30.f4 c4 31.Bg2 Re2
  =  (-0.23)   Depth: 14   00:00:00  1kN
24.Nc6+ Bxc6 25.Qxd6 Rxd6 26.dxc6 Rxc6 27.Re7+ Kb6 28.Rxf7 Re8 29.Kc1 Rd6 30.f4 c4 31.f5 gxf5 32.Bxf5 h5 33.Rd2 Re1+ 34.Rd1 Re5 35.Bc8 Re2
  =/+  (-0.45)   Depth: 15   00:00:15  5161kN
24.Nc6+ Bxc6 25.Qxd6 Rxd6 26.dxc6 Kb6 27.Re7 Kxc6 28.Rxf7 Re8 29.Bf1 Re3 30.a4 Kb6 31.axb5 axb5 32.f4 c4
  =/+  (-0.44)   Depth: 16   00:00:56  16417kN
24.Nc6+ Bxc6 25.Qxd6 Rxd6 26.dxc6 Kb6 27.Re7 Kxc6 28.Rxf7 Re8 29.Bf1 Re3 30.f4 Kb6 31.c3
  =/+  (-0.46)   Depth: 17   00:01:05  19343kN
24.Rxd4 Kb6 25.b4 Qxf4 26.Rxf4 Nxd5 27.Rxf7 cxb4 28.axb4 Nxb4 29.Nb3 Rd6 30.Re6 Rhd8 31.Rff6 Rxe6 32.Rxe6+ Bc6 33.c3 Nd5 34.Kc2 Re8 35.Rxe8 Bxe8 36.Kd3 a5 37.Nd4
  =/+  (-0.40)   Depth: 17   00:01:27  26255kN
24.Rxd4 Kb6 25.b4 Qxf4 26.Rxf4 Nxd5 27.Rxf7 cxb4 28.axb4 Nxb4 29.Nb3 Rd6 30.Re6 Rhd8 31.Rff6 Rxe6 32.Rxe6+ Bc6 33.Kb2 Re8 34.Rxe8 Bxe8 35.Be6 Bc6 36.Nd2 a5 37.c4
  =/+  (-0.37)   Depth: 18   00:02:27  43286kN

(, Microsoft 15.11.2008)

Never under estimate the great Rybka
Parent - - By turbojuice1122 (Gold) [us] Date 2008-11-14 21:25
Yes--I would guess that Deep Fritz 11 was tuned to find that move in that position.
Parent - - By Uly (Gold) [mx] Date 2008-11-15 01:42
Then someone may try reversing colors/mirroring to see what happens.
Parent - - By turbojuice1122 (Gold) [us] Date 2008-11-15 01:46
No, I don't think that the position was literally programmed into the engine--I think that there is a good possibility that some parameters were tuned to obtain the best result for that position among many others.
Parent - - By Dragon Mist (****) [hr] Date 2008-11-15 02:16
I do expect Deep Fritz 11 to be deep, but not THAT deep. :-)
Parent - By turbojuice1122 (Gold) [us] Date 2008-11-15 10:19
That wouldn't really be too surprising, as many programmers use that type of method, from what I hear.  How effective is the method remains to be seen.
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka can't find 24.Rxd4!

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill