Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / what is the minimal evaluation by rybka when humans resigned
- - By Uri Blass (*****) Date 2008-01-25 15:50
when they had a justification to resign.

I ask it because I found that in one of kasparov's games against karpov
kasparov resigned when rybka's evaluation even after some hours of search is only -1.28 and I think that 1.28 pawns of rybka is not a case when the position is hopeless.

I posted the relevant position with analysis by rybka in the chess forum because I was not sure if the discussion about specific position is about rybka or about chess and my question about the reason that kasparov resigned is about chess and not about rybka because it is not a case that I see a justified reason for kasparov to resign.

Uri
Parent - - By turbojuice1122 (Gold) Date 2008-01-25 18:16
In the game to which you're referring, it could be that Kasparov thought he saw something "obvious" that wasn't really there.  It has happened before and since.  I'm sure that if you leave Rybka on for awhile in the final position in which Kasparov resigned in Game 2 of the 1997 Deep Blue match, it will show 0.00.  Grandmasters have definitely resigned before in winning positions--Chessbase occasionally has a story on such a thing perhaps on the order of once or twice per year, though I don't remember any specific games.
Parent - - By Uri Blass (*****) Date 2008-01-25 18:36
Note that in the kasparov-karpov game kasparov resigned after analysis of the adjourned game.

My question is about resigning in lost positions and not about resigning in drawn positions so the game against deeper blue is not relevant.
The game against deeper blue is not relevant for another reason and kasparov resigned only because he overestimated deeper blue and assumed that there is no chance that it can miss a perpetual check so he even did not analyze that possibility.

My question is what is the minimal evaluation by rybka when chess players resigned when the decision is justified.

Uri
Parent - - By M ANSARI (*****) Date 2008-01-25 20:39
I have GM games where people resigned in drawn and even in won positions.  But these would be considered blunders and I think you mean positions where a GM resigned even though the evaluation is not so bad.  I think there are many positions where even though there is a slight advantage for one side, the horizon effect is such that a computer would not know how the game will be won but will win it anyway ... while a human is familiar with the position and can see no way to defend.  Usually that is when he is locked in a very passive position where his pieces are tied up taking care of defensive tasks while the other side has the ability to do something else (like 3 pawns vs 2 pawns on the other flank which will obviously Queen).
Parent - By turbojuice1122 (Gold) Date 2008-01-26 06:05
I now get the impression he is talking more about a Rybka issue than a GM issue--he is specifying now that the requirement is that the GM's decision be justified, so we're practically talking about situations in which the GM's evaluation that the position is lost is better than Rybka's...(I think)...
Parent - - By Mark (****) Date 2008-01-25 20:37

> I'm sure that if you leave Rybka on for awhile in the final position in which Kasparov resigned in Game 2 of the 1997 Deep Blue match, it will show 0.00.


I though this was impossibly deep for computers to find, due to some quiet moves in the middle of the drawing sequence?  That would be pretty amazing if Rybka could find it.  I'll let it run for awhile to see.
Parent - - By George Tsavdaris (****) Date 2008-01-25 22:28
Rybka 1.2i 32 bit on a AMD 3500+ 2.2GHz finds Qe3 quickly and gives a score of about 0.32 after some time.
I guess Rybka 2.3.2a on a QUAD would do extremely better and even give a 0.00 score.

Analysis by Rybka 1.2i 32-bit:

45...Qb6xc6 46.d5xc6 Rb8-c8 47.Ra6-a7+ Rc8-c7 48.Ra7-a5 Kf7-e7 49.Kf1-e2 Rc7-c8 50.Ra5xb5 Bd6-c7 51.Rb5-d5 Rc8-d8 52.b4-b5
  +-  (1.58)   Depth: 15   00:00:06  629kN
45...Qb6xc6 46.d5xc6 Rb8-c8 47.Ra6-a7+ Rc8-c7 48.Ra7-a5 Kf7-e7 49.Kf1-e2 Rc7-c8 50.Ra5xb5 Bd6-c7 51.Rb5-d5 Rc8-d8 52.b4-b5
  +-  (1.49)   Depth: 16   00:00:07  934kN
45...Qb6xc6 46.d5xc6 Rb8-c8 47.Ra6-a7+ Rc8-c7 48.Ra7-a5 Kf7-e7 49.Kf1-e2 Rc7-c8 50.Ra5xb5 Bd6-c7 51.Rb5-d5 Rc8-d8 52.b4-b5
  +-  (1.60)   Depth: 17   00:00:08  1491kN
45...Qb6xc6 46.d5xc6 Rb8-c8 47.Ra6-a7+ Rc8-c7 48.Ra7-a5 Kf7-e7 49.Ra5-a8 Ke7-f7 50.Kf1-e2 Bd6-f8 51.Ra8-a5 Bf8-d6 52.Ra5xb5
  +-  (1.75)   Depth: 18   00:00:15  3273kN
45...Qb6xc6 46.d5xc6 Rb8-c8 47.Ra6-a7+ Rc8-c7 48.Ra7-a8 h6-h5 49.Kf1-e2 Kf7-e7 50.Ra8-g8 Ke7-f7 51.Be4-d5+ Kf7-e7 52.Ke2-e3
  +-  (1.86)   Depth: 19   00:00:27  6445kN
45...Qb6-e3 46.Qc6xd6 Rb8-e8 47.h3-h4 h6-h5 48.Be4-f3 Qe3-c1+ 49.Kf1-f2 Qc1-d2+ 50.Bf3-e2 Qd2-f4+ 51.Kf2-g1 Qf4-e3+ 52.Kg1-h2
  +/-  (0.77)   Depth: 19   00:02:18  17786kN
45...Qb6-e3 46.Qc6xd6 Rb8-e8 47.h3-h4 h6-h5 48.Be4-f3 Qe3-c1+ 49.Kf1-f2 Qc1-d2+ 50.Bf3-e2 Qd2-f4+ 51.Kf2-g1 Qf4-e3+ 52.Kg1-h2
  +/=  (0.42)   Depth: 20   00:03:08  23328kN
45...Qb6-e3 46.Qc6xd6 Rb8-e8 47.h3-h4 h6-h5 48.Be4-f3 Qe3-c1+ 49.Kf1-f2 Qc1-d2+ 50.Bf3-e2 Qd2-f4+ 51.Kf2-g1 Qf4-e3+ 52.Kg1-h2
  +/=  (0.38)   Depth: 21   00:04:22  32681kN
45...Qb6-e3 46.Qc6xd6 Rb8-e8 47.Qd6-d7+ Re8-e7 48.Qd7-c6 Qe3xe4 49.d5-d6 Qe4xf5+ 50.Kf1-g1 Re7-d7 51.Qc6xb5 Qf5-e6 52.Qb5-c5
  +/=  (0.50)   Depth: 22   00:06:34  48034kN
45...Qb6-e3 46.Qc6xd6 Rb8-e8 47.Qd6-d7+ Re8-e7 48.Qd7-c6 Qe3xe4 49.d5-d6 Qe4xf5+ 50.Kf1-g1 Re7-d7 51.Qc6xb5 Qf5-e6 52.Qb5-c5
  +/=  (0.45)   Depth: 23   00:09:45  70642kN
45...Qb6-e3 46.Qc6xd6 Rb8-e8 47.Qd6-d7+ Re8-e7 48.Qd7-c6 Qe3xe4 49.d5-d6 Qe4-d3+ 50.Kf1-g1 Re7-e8 51.Ra6-a7+ Kf7-f8 52.Qc6-c5
  +/=  (0.50)   Depth: 24   00:14:50  107mN
45...Qb6-e3 46.Qc6xd6 Rb8-e8 47.Qd6-d7+ Re8-e7 48.Qd7-c6 Qe3xe4 49.d5-d6 Qe4-d3+ 50.Kf1-g1 Re7-e8 51.Ra6-a7+ Kf7-f8 52.Qc6-c5
  +/=  (0.42)   Depth: 25   00:26:33  184mN
45...Qb6-e3 46.Qc6xd6 Rb8-e8 47.Qd6-d7+ Re8-e7 48.Qd7-c6 Qe3xe4 49.d5-d6 Qe4-d3+ 50.Kf1-g1 Re7-e8 51.Ra6-a7+ Kf7-f8 52.Qc6-c5
  +/=  (0.32)   Depth: 26   00:51:10  353mN
Parent - - By Mark (****) Date 2008-01-26 13:57
I let it run overnight and after almost 3 billion nodes and 29 ply it still doesn't see a forced draw.

45...Qe3 46.Qxd6 Re8 47.Qc7+ Re7 48.Qc6 Qxe4 49.d6 Qd3+ 50.Kg1 Re8 51.Ra1 Qxc3 52.Rd1
  +/=  (0.29)

I'll force the first few moves and run it again, but I still think an actual draw score of 0.00 is out of reach.
Parent - By Uri Blass (*****) Date 2008-01-26 14:27
Of course it is out of reach because no forced draw is proved.
The only proved result is equal material with small positional advantage for white that is probably not enough to win.

Deeper blue clearly did not see it and could not avoid fast draw in case that kasparov was playing Qe3
but playing Qe3 is no easy draw for black against correct defence.

Uri
Parent - - By lkaufman (*****) Date 2008-01-26 05:38
There have been plenty of GM games when one player resigned in a position that was very bad, possibly lost but not clearly so. Ivanchuk in particular used to do this. Sometimes he just lost a pawn and was so upset that he resigned even though he might have had a 10% chance to save a draw. Anand recently resigned against Topalov when he had an ugly position but the win was not 100% clear. These are generally emotional decisions or sometimes someone resigns when there is still a small chance because the chance is not enough to justify possibly jeopardizing the next game by exhaustion and/or depression.
Parent - - By turbojuice1122 (Gold) Date 2008-01-26 06:07
Yes, I was about to bring up the game of which you speak between Anand and Topalov.  Anand's position did look hopeless, but I recall from analyzing at the time that Rybka gave Topalov a score of only about 0.5 or 0.6.
Parent - - By M ANSARI (*****) Date 2008-01-26 11:27
Yes but if you played through the game then you would quickly see that Anand was totally passive and Topalov's plan was simply to exhange rooks (which was forced) and thus the resulting game is obviously lost.  I sometimes like to take positions that someone has resigned and try the same position and try to win the game against Rybka playing blitz.  I was mortified to find out how much I suck at the game of chess since on many occasions I failed to win and in some cases even lost.  But again if you have played through the entire game and was involved in the struggle you can see when you have no more tricks and thus can give up.  I mean any of us here would resign if faced with K vs K and rook ... for GM's this would be as if they were facing a similar thing.
Parent - By Uly (Gold) Date 2008-01-26 18:34

> I mean any of us here would resign if faced with K vs K and rook


Not me! At my level, when I have my lone King against King+Rook, I still have a chance of drawing in time (If my opponent used most of its time to win, but now has to move too fast) or drawing by 50 move rule or 3 time repetition or even stalemate. It has happened, so till I become stronger, my slogan will be "Never resign!" (Though, against much stronger opponents than me, I sometimes resign after just losing a pawn, as it's very apparent that my opponent isn't going to do anything to let me at least draw.)
Parent - By molot Date 2008-08-10 18:29

> "Yes but if you played through the game then you would quickly see that Anand was totally passive and Topalov's plan was simply to exhange rooks (which was forced) and thus the resulting game is obviously lost."


I agree that the final position was objectively lost but Topalov himself pointed out that he has several times in recent years failed to win - or even lost - positions that were more clearly winning than that...

Kasparov also thought the resignation was premature.

Anand was probably just sick of the position, which is understandable, and being in such a mood does not increase your chances of saving a difficult game...
Up Topic Rybka Support & Discussion / Rybka Discussion / what is the minimal evaluation by rybka when humans resigned

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill