Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Rybka's oddity! Can anyone reproduce this......?

Author: Uri Blass

Date: 14:52:20 02/02/06

Go up one level in this thread


On February 02, 2006 at 17:44:30, Uri Blass wrote:

>On February 02, 2006 at 16:07:20, George Tsavdaris wrote:
>
>> I noticed a strange behaviour and definitely a bad one from latest Rybka, and
>>for Winfinder so that means it should be a general bad behaviour of Rybka
>>family:-)  Possible bug that prevents correct analysis.....
>>
>> This behaviour is noticed at Fritz 8 GUI(i haven't tried Arena yet). It is also
>>happening at Deep Fritz 8 GUI. It must be a hashproblem or something.....
>> Conditions(I'll describe the procedure i followed--Keep in mind that the values
>>for the seconds i've made analysing between the moves should not play any role
>>at all and different values would produce the problem again):
>>
>>Set the following game:
>>
>>[Event "??"]
>>[Site ""]
>>[Date "2005.01.03"]
>>[Round "?"]
>>[White "Tsavdaris"]
>>[Black "????"]
>>[Result "1-0"]
>>[PlyCount "54"]
>>[EventDate "2005.??.??"]
>>
>>1. e4 e5 2. f4 d5 3. exd5 exf4 4. Nf3 Nf6 5. c4 c6 6. Nc3 cxd5 7. d4 Bb4 8. Be2
>>Ne4 9. O-O Bxc3 10. bxc3 Nxc3 11. Qc2 Nxe2+ 12. Qxe2+ Be6 13. Ba3 dxc4 14. d5
>>Qxd5 15. Rad1 Qb5 16. Nd4 Qb6 17. Kh1 Nd7 18. Rxf4 O-O-O 19. Nxe6 Qxe6 20.
>>Rxc4+ Kb8 21. Qxe6 fxe6 22. Rcd4 b5 23. Rxd7 Rxd7 24. Rxd7 Rc8 25. Bd6+ Ka8 26.
>>Bc7 a6 27. Rxg7 Rf8 1-0
>>
>>-Set Rybka as an engine, clear hashtables and set it to let's say 256 MB.
>>-Go to move 14 for black and start analysis with Rybka(multivariation with 4
>>lines i had) for some seconds(4-5).
>>-It displays the move Rad1 as best. Press the "Right Key" button from the
>>keyboard in order the next move to be executed while the analysis will continue.
>>-Press the "Right Key" button again after 1-2 seconds so the move Qb5 will be
>>played.
>>-Press the "Right Key" button again after 4-5 seconds so the move Nd4 will be
>>played.
>>-Press the "Right Key" button again after 1-2 seconds.
>>Now Rybka displays 17.Rxf4 as the best move with -2.00 evaluation!
>>So what? Well if we clear at that point the hashtables then Rybka displays the
>>correct evaluation and correct move which is 17.Kh1 with a positive score.
>>
>> Also if we put the position with white at move 17 to play, to Rybka to analyse
>>then we get the correct analysis. But if we have the analysis open from move 15
>>and continue up to move 17 for white we get 17.Rxf4????
>>
>>
>>Can anyone reproduce this.....?
>
>I could reproduce it in 3 option mode
>
>Here is the pv at depth 14
>
>Tsavdaris,  2005
>rn2k2r/pp3ppp/1q2b3/8/2pN1p2/B7/P3Q1PP/3R1RK1 w kq - 0 1
>
>Analysis by Rybka 1.01 Beta 13b 32-bit:
>
>1. -+  (-2.38): 17.Rxf4
>2. -+  (-2.38): 17.Kh1
>3. -+  (-2.53): 17.Bb2
>
>(,  03.02.2006)
>
>depth 15:
>
>Tsavdaris,  2005
>rn2k2r/pp3ppp/1q2b3/8/2pN1p2/B7/P3Q1PP/3R1RK1 w kq - 0 1
>
>Analysis by Rybka 1.01 Beta 13b 32-bit:
>
>1. -+  (-2.44): 17.Rxf4
>2. -+  (-2.44): 17.Kh1
>3. -+  (-2.59): 17.Bb2
>
>(,  03.02.2006)
>
>one of the strenge things is that there is no pv for none of the lines
>
>Uri

A simpler way to reproduce it is to start from the position after 15.Rad1 and
make the game moves 15...Qa5 16.Nd4 Qb6 fast.

I get the following analysis at depth 14:

Tsavdaris,  2005
rn2k2r/pp3ppp/1q2b3/8/2pN1p2/B7/P3Q1PP/3R1RK1 w kq - 0 1

Analysis by Rybka 1.01 Beta 13b 32-bit:

1. -+  (-2.44): 17.Bb2
2. -+  (-2.45): 17.Rxf4
3. -+  (-2.45): 17.Kh1

(,  03.02.2006)




Uri



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.