Author: blass uri
Date: 01:59:23 07/08/99
Go up one level in this thread
On July 08, 1999 at 03:37:53, Paulo Soares wrote:
>On July 07, 1999 at 20:24:55, Robin Smith wrote:
>
>>On July 03, 1999 at 11:43:37, blass uri wrote:
>>
>>>
>>>On July 02, 1999 at 17:06:44, Robin Smith wrote:
>>>
>>>>On July 02, 1999 at 06:53:52, blass uri wrote:
>>>>
>>>>>
>>>>>On July 02, 1999 at 03:30:04, Robin Smith wrote:
>>>>>
>>>>>>On July 01, 1999 at 23:49:48, Ted Sutton wrote:
>>>>>>
>>>>>>>Robin,
>>>>>>> Thanks for a fascinating post with regard to positions in the 11th USCCC where
>>>>>>>you believe your opponents relied on a computer for a faulty evaluation.
>>>>>>> This would be a very interesting and relevant test case for us,if you would
>>>>>>>care to share several of these positions with us (naturally without names).
>>>>>>>We are dealing here with very strong correspondence play, so can see what of
>>>>>>>errors of positional judgement computers make.
>>>>>>> (I was unaware that computer use is legal in the USCCC, but it is logical,
>>>>>>>since the USCCC is an ICCF sponsored event, and computers are legal under ICCF).
>>>>>>> In view of the fact that computer use was legal in these games, then your
>>>>>>>opponents, assuming they were consulting with computers, were acting ethically
>>>>>>>and legally (though perhaps unwisely), and there is no reason not to disclose
>>>>>>>these positions.
>>>>>>
>>>>>>Sure, here are a couple. Of course I don't KNOW how my opponents generated
>>>>>>moves, it's just a theory. Interestingly the errors are more tactical than
>>>>>>positional, but the lines are pretty deep. The 1st one really looks computer
>>>>>>horizon effect. The 2nd game was published in the November 1998 Chess. Modern
>>>>>>computers/programs probably do better than at the time these games were played.
>>>>>>
>>>>>>[Event "US11F"]
>>>>>>[Date "1995-1998"]
>>>>>>[White "Smith, Robin"]
>>>>>>[Black "Thompson, Paul"]
>>>>>>[Result "1-0"]
>>>>>>[ECO "B89"]
>>>>>>[Annotator "Robin Smith"]
>>>>>>
>>>>>>1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 Nc6 6. Bc4 e6 7. Be3 Be7 8.
>>>>>>Qe2 a6 9. O-O-O O-O 10. Bb3 Qc7 11. g4 Nxd4 12. Rxd4 Nd7 13. g5 Nc5 14. Rg1 b5
>>>>>>15. e5 dxe5 16. Rh4 Nxb3+ 17. axb3 g6 18. Qf3!? {White gambles on a strong
>>>>>>sacrificial attack that should be a draw with correct defense by Black.} Bb7 19.
>>>>>>Qh3 h5 20. Rxh5 gxh5 21.Qxh5 Bf3??
>>>>>>
>>>>>>Black is now lost. Computers like this move because, being up a whole rook,
>>>>>>they don't like to settle for a draw. 21... Bc5! is the only move after which
>>>>>>22. Bxc5 Qxc5 23. g6 Kg7!=draw
>>>>>
>>>>>I do not think that computers like this move if you give them many hours.
>>>>>I tried the position with Junior5.4 and it found Bc5 in less than 10 minutes.
>>>>>I do not know the exact time because I did other things at the same time and the
>>>>>computer may be slowed down.
>>>>>It evaluates Bc5 as 0.52 pawns for white and gives the same line
>>>>>22.Bxc5 Qxc5 23.g6 Kg7...
>>>>>
>>>>>Uri
>>>>
>>>>The move 21. ...Bf3?? was played several years ago. At that time computers were
>>>>not as good as today.
>>>
>>>I gave Genius3 on pentium100 the position and after 87 minutes and 2 seconds it
>>>does not like Bf3(evaluation -0.36) but expects Qxf3 instead of Qh6.
>>>
>>>It is now considers after 2 hours Bc5.
>>>
>>> Also, after 21. ...Bc5 22.Bxc5 Qxc5 23.g6 Kg7 I am
>>>>surprised Junior5.4 gives a White advantage. I believe that the best plan for
>>>>White is to 3 repeat the position with perpetual checks. What does Junior 5.4
>>>>give as White's best continuation after you force the moves through 23. ...Kg7?
>>>
>>>white can have a strong attack and Junior evaluates the attack+the bishop at b7
>>>as better than the rook.
>>>I think that Junior is correct in this evaluation.
>>>I gave it to play against itself(time control 270 minutes for 40 moves) and it
>>>played
>>>24.Qh7+ Kf6 25.Qh4+ Kg7 26.b4 Qb6 27.gxf7+ Kxf7 28.Qh7+ Ke8 29.Rg7 Qd6
>>>30.Rxb7 Rxf2 31.b3 e4 32.Qg6+ Kd8 33.Nxe4 Qf4+ 34.Kb2 Rxc2+
>>>with evaluation 4.10 pawns for white.
>>>(it saw 3.98 pawns for white when it played 33.Nxe4)
>>>
>>>Uri
>>
>>Interesting. I went into this whole line figuring I had at least a draw, and
>>after 21 ... Bf3? I never looked any further at possible white wins. Now I am
>>intrigued. I think I will analyze this again. If white can force a win against
>>all defenses this game could have real opening signifigance.
>>
>>You also mention best hardware/software at the time could have found 21. ...
>>Bc5. You are probably correct, but I obviously don't "know" a computer was
>>being used, just that the move Bf3 seems computer like to me. Also, it is still
>>possible that my opponant was using a computer but not the "best"
>>hardware/software combination available at the time.
>
>Robin, I made some analyses based in the line presented for Uri, with
>the aid of Fritz5.32, will be that blacks obtain to draw?
>Instead of 26... Qb6, I believe that 26... Qc4 can be better for black,
>look:
>
>26... Qc4 27. gxf7+ (27. Rg4 Qf1+ 28. Nd1 Bf3 29. Qh7+ (29. gxf7+ Bxg4 - +)
>29... Kf6 30. Qh4+ =) 27... Kxf7 28. Qh7+ Ke8 29. Qxb7 (29. Rg7
>Bf3 30. Re7+ Kd8 31. Rd7+ Kc8 32. Qg7 Qf4+ 33. Kb1 Qf6 - +) Rd8 unclear
>
>In the correspondence game, the player who has the aid of the computer,
>does not have to leave the program analyzing much time one move,
>the correct is to force the program to play one move chosen for the
>proper player and to follow the line.
I agree that it may be a better idea not to give the computer a long time for
one move and try to follow the main line but I am not sure if this example
proves it.
Junior5.4 used only 976 seconds on my pentium200 to choose 26...Qb6
I am not sure if it is going to choose Qb6 if I give it some hours per move.
If you do not give the computer some hours per move then
there is a chance that you can also miss a surprising move that the computer can
find after some hours but I believe that you can earn more then you lose by
trying to follow some lines and not only giving the computer a long time in the
root position.
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.