Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Program bug or Autoplayer woes?

Author: Christophe Theron

Date: 08:54:32 11/12/99

Go up one level in this thread


On November 12, 1999 at 04:01:39, Bertil Eklund wrote:

>On November 12, 1999 at 02:00:32, Ed Schröder wrote:
>
>>On November 11, 1999 at 20:18:18, Howard Exner wrote:
>>
>>>Here's the game which ended in a draw.[Event "SSDF 40-2"]
>>>[Site "DK-Albertslund"]
>>>[Date "1999.11.10"]
>>>[Round "23"]
>>>[White "Chess Tiger 12.0 dos K6-2 450,64"]
>>>[Black "Shredder 2 P200MMX 24+8"]
>>>[Result "1/2-1/2"]
>>>[WhiteElo "2645"]
>>>[BlackElo "2503"]
>>>[Annotator "Test: HC Lykke"]
>>>
>>>1. e4 c5 2. Nc3 d6 3. f4 Nc6 4. Nf3 Nf6 5. Bb5 e6 6. Bxc6+
>>>bxc6 7. d3 Be7 8. e5 Nd5 9. Ne4 dxe5 10. Nxe5 Bb7 11. O-O
>>>O-O 12. Qe2 f6 13. Nc4 Ba6 14. Bd2 Nb6 15. Na5 Qd7 16. Be3
>>>Na4 17. Qf2 c4 18. b3 f5 19. Nd2 cxd3 20. bxa4 Bb4 21. Nab3
>>>dxc2 22. Rfe1 Rad8 23. a3 Bc3 24. Nc5 Qe7 25. Nxa6 Bxa1 26.
>>>Rxa1 Qxa3 27. Qe1 Qc3 28. Rc1 Rd3 29. Kf2 Qb2 30. Nc4 Qa2
>>>31. Ne5 Rd1 32. Qc3 Rxc1 33. Bxc1 Qxa4 34. Nc5 Qa2 35. Ncd7
>>>Re8 36. Qxc6 Rd8 37. Qc7 Ra8 38. Qb7 Rf8 39. Nxf8 Qa5 40.
>>>Qf7+ Kh8 41. Nxe6 Qb6+ 42. Kg3 Qb3+ 43. Kf2 Qb6+ 44. Ke2
>>>Qa6+ 45. Kf2 Qb6+ 1/2-1/2
>>>
>>>Would the game end the same from move 41 on if it were manually played?
>>
>>Rebel-Tiger (beta) plays of course 43.Kh4 announcing mate in 3.
>>
>>No 45.Kf2??? also but 45.Ke3 announcing mate in 5.
>>
>>Perhaps something is wrong with the PGN moves?
>>
>>Even in the end-position Rebel-Tiger (beta) still announces a
>>mate.
>>
>>Ed
>Hi!
>
>In one of my testgames Tiger used 30(!!) moves for a simple mate in 4 or 5
>moves. Tiger won the game but  after 10-15 moves with the same announcement mate
>in 4 or 5 I began to wonder if it should be a draw according to the 50-moves
>rule.
>
>Bertil SSDF


I have to say my 'mea culpa' here.

The problem was not the engine in this case but the time manager.

I have noticed during the French Championship last year that Tiger could spend a
long time before playing an obvious mate. In fact a mate move was considered as
any other move, so Tiger took his time to play it. This means probably one or
two minutes in a long time controls game.

I thought that this was very bad manners for the opponent, so I decided to make
it more "elegant" and change the time management.

The new version stopped to search as soon as a mate in X move was found. So in
this case it moved instantly.

Great. But it is a bloody mistake. What happens? First, Tiger finds a mate in,
say, 5 moves. So it stops and announces "mate in 5". Then, on the next move, the
first mate it finds is a mate in 6. So it stupidly stopped and announced a mate
in 6, and was happy with it.

This could be an infinite cycle, Tiger announcing mates in several moves, but
never actually checkmating the opponent.

I have discovered this (old) bug on Nov 2, 1999, and it is now FIXED. The
current beta version does NOT have the bug, and that's why Ed cannot reproduce
it.

Thanks to Howard Exner for pointing this out.



    Christophe



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.