Author: Uri Blass
Date: 03:15:29 01/05/05
Go up one level in this thread
On January 05, 2005 at 06:03:39, Uri Blass wrote:
>On January 05, 2005 at 05:54:19, Sedat wrote:
>
>>On January 05, 2005 at 05:41:52, Kurt Utzinger wrote:
>>
>>>Hi Sedat
>>>Were conditions indeed identical. I doubt because in some games
>>>the first move was 1.Sbd2 in other games 9.Sbd2 and this may
>>>have an influence on the way the engines are using the time
>>>management.
>>>Kind regards
>>>Kurt
>>
>>Hi Kurt,
>>
>>For all GUI's the nunn position was same :
>>
>>Just under ChessPartner GUI :Gandalf 6.0 UCI did not want to play with :
>>
>>[FEN "rn1qk2r/p3bppp/bpp1pn2/3p4/2PP4/1PB2NP1/P3PPBP/RN1QK2R w KQkq d6 0 9"]
>>
>>and thats why i decided to create a nunn position under ChessPartner GUI:
>>
>>1. d4 Nf6 2. Nf3 e6 3. c4 b6 4. g3 Ba6 5. b3 Bb4 6. Bd2 Be7 7. Bg2 c6 8. Bc3 d5
>>
>>Best,
>>Sedat
>
>It means different conditions even without the time management problem.
>
>Getting Fen and getting moves is not the same because the order of pieces in the
>memory of the computer may be different.
>
>Note also that games may be different with movei even with the same conditions.
>
>The problem is that if movei does not stop at the end of the iteration then it
>may ponder in one game for 8192*n nodes and in another game for 8192*(n+1) nodes
>because it does not count time accurately(it may also stop at different
>iteration because of the same problem but it is more rare problem).
>
>Different number of nodes may cause different order of pieces in the array and
>even if it does not cause immediately different move it may cause a different
>move later in the game.
>
>I guess that this problem is mainly in games with ponder on because with ponder
>off movei almost always stop at the end of the iteration.
>
>Uri
I see that you use ponder off.
With ponder off movei more rarely stops at the middle of the iteration
but if it does it then it may be a problem that can cause different games(even
under the same gui).
Not countin time correctly may cause it to stop at different iteration in
different games.
With ponder on counting time is not the problem because movei stops at the
middle of the iteration almost only when the opponent plays an expected but the
opponent may not use exactly the same time so movei may search 8192*n nodes when
it finds that the opponent plays a move in game 1 and 8192*(n+1) nodes when it
finds that the opponent plays a move in game 2.
It caused it to avoid doubles even with learning.
In WBEC(ponder on) movei learned to repeat the same opening that helped it to
win but failed to repeat move 23 probably because of that problem.
[Event "WBEC9 1st Division"]
[Site "MP2600"]
[Date "2005.01.01"]
[Round "10.2"]
[White "Movei 0.08.295"]
[Black "Abrok 5.0"]
[Result "1-0"]
[PlyCount "84"]
[EventDate "2005.??.??"]
[TimeControl "40/2400:0"]
1. d4 d5 2. c4 c6 3. Nc3 dxc4 4. e4 b5 5. a4 b4 6. Na2 e5 7. Nf3 exd4 8. Bxc4
Ba6 9. Qb3 Qe7 10. O-O c5 11. Bd5 Bb7 12. Ne5 Bxd5 13. Qxd5 Nf6 14. Qxa8 Qxe5
15. g3 Nh5 16. Qxa7 Bd6 17. Qb6 O-O 18. f4 Qe6 19. e5 Qg4 20. Qb7 Be7 21. Bd2
Nd7 22. a5 Rb8 23. Qe4 Qe6 24. Qg2 g6 25. g4 Ng7 26. f5 gxf5 27. gxf5 Qxe5 28.
a6 Rc8 29. Rae1 Qf6 30. a7 Bf8 31. a8=Q Rxa8 32. Qxa8 b3 33. Qa4 bxa2 34. Qxd7
c4 35. Ra1 Qe5 36. Rxa2 Qe4 37. Ra7 Qg4+ 38. Kf2 Qh4+ 39. Kf3 Qh5+ 40. Kg3 h6
41. Bxh6 d3 42. Bd2 Kh7 {Abrok resigns} 1-0
[Event "WBEC9 1st Division"]
[Site "MP2600"]
[Date "2005.01.01"]
[Round "10.4"]
[White "Movei 0.08.295"]
[Black "Abrok 5.0"]
[Result "1-0"]
[PlyCount "76"]
[EventDate "2005.??.??"]
[TimeControl "40/2400:0"]
1. d4 d5 2. c4 c6 3. Nc3 dxc4 4. e4 b5 5. a4 b4 6. Na2 e5 7. Nf3 exd4 8. Bxc4
Ba6 9. Qb3 Qe7 10. O-O c5 11. Bd5 Bb7 12. Ne5 Bxd5 13. Qxd5 Nf6 14. Qxa8 Qxe5
15. g3 Nh5 16. Qxa7 Bd6 17. Qb6 O-O 18. f4 Qe6 19. e5 Qg4 20. Qb7 Be7 21. Bd2
Nd7 22. a5 Rb8 23. Qg2 Rf8 24. a6 Nb6 25. a7 Na8 26. Qb7 Re8 27. f5 h6 28. Rf2
b3 29. Qxb3 Qe4 30. f6 gxf6 31. Qa4 Rc8 32. Qd7 Rc7 33. Qe8+ Kg7 34. Raf1 Qc6
35. Bxh6+ Kg6 36. Qg8+ Kxh6 37. Qh8+ Kg6 38. g4 Nf4 {Abrok resigns} 1-0
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.