Author: Robert Hyatt
Date: 13:30:14 01/16/04
Go up one level in this thread
On January 16, 2004 at 15:21:34, Bob Durrett wrote:
>On January 16, 2004 at 14:42:15, Robert Hyatt wrote:
>
>>On January 16, 2004 at 14:07:11, Drexel,Michael wrote:
>>
>>>On January 16, 2004 at 13:14:18, Robert Hyatt wrote:
>>>
>>>>On January 15, 2004 at 23:20:02, Drexel,Michael wrote:
>>>>
>>>>>On January 15, 2004 at 20:41:15, Robert Hyatt wrote:
>>>>>
>>>>>>On January 15, 2004 at 17:07:01, Drexel,Michael wrote:
>>>>>>
>>>>>>>On January 15, 2004 at 16:27:13, Robert Hyatt wrote:
>>>>>>>
>>>>>>>>On January 15, 2004 at 15:09:07, Geoff wrote:
>>>>>>>>
>>>>>>>>>Hi
>>>>>>>>>
>>>>>>>>>I suspect Mr Kasparov would probably want a few dollars before considering that
>>>>>>>>>option.
>>>>>>>>>It made me wonder who is the strongest human opponent Crafty has played against?
>>>>>>>>>I would be intrigued to know what the match score was ?
>>>>>>>>>
>>>>>>>>> Geoff
>>>>>>>>>
>>>>>>>>>PS
>>>>>>>>>Thanks to Prof. Hyatt for the links to the online papers. Some interesting
>>>>>>>>>reading material there.
>>>>>>>>
>>>>>>>>
>>>>>>>>It's played Shirov a lot of blitz. Won most. It's played Karpov some
>>>>>>>>games and won them all but the sample size was small and it was a couple of
>>>>>>>>years ago. It has played Yasser, Benjamin, Kamsky, I'd hate to try to name
>>>>>>>>them all...
>>>>>>>>
>>>>>>>>It does very well at blitz, winning probably 90%. At game/30 it wins way
>>>>>>>>more than it loses. Beyond that, who knows...
>>>>>>>
>>>>>>>Has anyone ever beated Crafty in such a convincing way?
>>>>>>>The game was played immediately after reading your post. I saw no engine output
>>>>>>>and didn't analyze the opening before the game.
>>>>>>>But it helped a lot to know that Crafty will never take on g5. People on ICC
>>>>>>>who don't read CCC do not know this.
>>>>>>>
>>>>>>>Michael
>>>>>>>
>>>>>>>
>>>>>>>[Event "Blitz:10'+10""]
>>>>>>>[Site "AMD 1.6 Ghz"]
>>>>>>>[Date "????.??.??"]
>>>>>>>[Round "?"]
>>>>>>>[White "Drexel"]
>>>>>>>[Black "Crafty 19.08"]
>>>>>>>[Result "1-0"]
>>>>>>>[PlyCount "47"]
>>>>>>>[TimeControl "600+10"]
>>>>>>>
>>>>>>>{64MB, Shredder8.ctg, ATHLONXP} 1. c3 {0} b6 {0} 2. d4 {3} Bb7 {0} 3. Nf3 {3}
>>>>>>>e6 {0.51/12 57} 4. Bg5 {(e3) 7} Nf6 {0.22/12 40} 5. Nbd2 {6} Be7 {0} 6. e3 {
>>>>>>>(h4) 137} c5 {0} 7. Bd3 {(h3) 4} O-O {0} 8. Qe2 {(e4) 14} h6 {0.24/11 30} 9. h4
>>>>>>>{(Bh4) 26} cxd4 {0.52/10 29} 10. exd4 {3} Qc7 {0.61/10 26} 11. O-O-O {11} Nc6 {
>>>>>>>0.52/10 16} 12. Kb1 {6} Rac8 {0.60/10 22} 13. g4 {(Bxf6) 50} Nxg4 {0.25/10 28}
>>>>>>>14. Rdg1 {(Rhg1) 29} h5 {0.25/10 27} 15. Nh2 {(Rg2) 132} f6 {-0.22/10 35} 16.
>>>>>>>Nxg4 {(Be3) 28} hxg4 {0.69/10 50} 17. Qxg4 {23} f5 {1.55/9 5} 18. Qh5 {(Qg2) 87
>>>>>>>} Rf7 {1.81/10 34} 19. Nf3 {26} Bd6 {3.51/11 70} 20. Bc1 {(Be3) 90} Ne7 {
>>>>>>>3.91/10 53} 21. Ng5 {4} g6 {8.49/10 17} 22. Nxf7 {9} Bg3 {9.04/10 13} 23. Rxg3
>>>>>>>{(Nh6+) 13} Kxf7 {12.02/10 64} 24. Qh7+ {5} 1-0
>>>>>>
>>>>>>
>>>>>>It happens. But in long and serious games, the anti-trojan stuff is always
>>>>>>turned off by me, manually. It is really important for blitz games, where
>>>>>>the depth is not sufficient to avoid the mate when it is possible... On fast
>>>>>>hardware, I'd probably turn it off all the time, in fact... Except for bullet
>>>>>>games which is really the only place it had trouble with this before (2 1 games
>>>>>>for example).
>>>>>
>>>>>I don't think my attack in this game was unsound.
>>>>
>>>>Neither do I.
>>>>
>>>>>Crafty never had the chance to take that bishop.
>>>>
>>>>
>>>>I didn't look at that very carefully, I just took your comment to be
>>>>directed at the Trojan Horse code, and I wanted to point out that at
>>>>normal time controls, that is turned off, because it can easily see
>>>>the difficulties when it has enough time. This attack really only works
>>>>at bullet-type time controls on the kind of hardware I normally use on ICC.
>>>>
>>>>
>>>>>Knowing it will probably never take makes it just easier for me since I don't
>>>>>have to calculate those lines.
>>>>>
>>>>>[D] rn1q1rk1/pb1pbppp/1p2pn2/2p3B1/3P4/2PBPN2/PP1NQPPP/R3K2R b KQ - 0 8
>>>>>
>>>>>I delayed castling with Qe2 and Crafty played h6. That's a serious strategical
>>>>>mistake IMO. Chances are high that it would play h6 also on a Dual or Quad.
>>>>>
>>>>>You don't weaken your Kingside before White has castled short in this positions.
>>>>>Either you play h6 _before_ castling yourself (so that white is forced to move
>>>>>that bishop) or you play it after white has castled short. You don't play it at
>>>>>all if white has castled long.
>>>>>
>>>>>Michael
>>>>
>>>>
>>>>On my machine, it drops h6 inside 30 seconds and it doesn't go back to it. IE
>>>>Crafty knows exactly what you said about making luft (h6) which only creates a
>>>>target for the opponent to shoot at.
>>>>
>>>>Here's the output on a dual xeon 2.8ghz with the current code:
>>>>
>>>> 7-> 0.62 0.37 1. ... h6 2. Bf4 g5 3. Be5 cxd4 4.
>>>> cxd4 d6 5. Bg3 <HT>
>>>> 8 1.06 0.49 1. ... h6 2. Bf4 cxd4 3. cxd4 Nc6 4.
>>>> O-O Nb4 5. Bc4 Rc8
>>>> 8 2.10 0.38 1. ... cxd4 2. exd4 h6 3. Bxf6 Bxf6
>>>> 4. O-O-O d5 5. Kb1 Nd7 6. Rhe1
>>>> 8-> 2.58 0.38 1. ... cxd4 2. exd4 h6 3. Bxf6 Bxf6
>>>> 4. O-O-O d5 5. Kb1 Nd7 6. Rhe1 (s=4)
>>>> 9 4.07 0.33 1. ... cxd4 2. exd4 Nd5 3. Bxe7 Qxe7
>>>> 4. O-O-O Nf4 5. Qf1 Bxf3 6. Nxf3 Qd6
>>>> (s=3)
>>>> 9-> 4.72 0.33 1. ... cxd4 2. exd4 Nd5 3. Bxe7 Qxe7
>>>> 4. O-O-O Nf4 5. Qf1 Bxf3 6. Nxf3 Qd6
>>>> 10 7.77 0.42 1. ... cxd4 2. exd4 h6 3. Bh4 Nc6 4.
>>>> O-O d5 5. Rfe1 Bd6 6. Bg3 Rc8 7. Bxd6
>>>> Qxd6
>>>> 10 14.65 0.33 1. ... h6 2. Bh4 d6 3. dxc5 bxc5 4.
>>>> O-O Nbd7 5. Rad1 Ng4 6. Bxe7 Qxe7 7.
>>>> e4
>>>> 10-> 14.65 0.33 1. ... h6 2. Bh4 d6 3. dxc5 bxc5 4.
>>>> O-O Nbd7 5. Rad1 Ng4 6. Bxe7 Qxe7 7.
>>>> e4 (s=2)
>>>> 11 22.42 0.43 1. ... h6 2. Bf4 cxd4 3. cxd4 Nc6 4.
>>>> O-O Nb4 5. Bb5 a6 6. Bc4 d5 7. Bb3
>>>> 11 42.98 0.26 1. ... cxd4 2. cxd4 Nd5 3. Bxe7 Qxe7
>>>> 4. O-O d6 5. Rac1 Nc6 6. Ne4 Rac8 7.
>>>> Nfg5 f6 <HT>
>>>> 11-> 44.37 0.26 1. ... cxd4 2. cxd4 Nd5 3. Bxe7 Qxe7
>>>> 4. O-O d6 5. Rac1 Nc6 6. Ne4 Rac8 7.
>>>> Nfg5 f6 <HT> (s=2)
>>>> 12 1:11 0.31 1. ... cxd4 2. exd4 Nd5 3. Bxe7 Qxe7
>>>> 4. O-O Nf4 5. Qe3 Nxd3 6. Qxd3 d6 7.
>>>> Rfe1 Nc6 8. Qe4
>>>> 12-> 1:19 0.31 1. ... cxd4 2. exd4 Nd5 3. Bxe7 Qxe7
>>>> 4. O-O Nf4 5. Qe3 Nxd3 6. Qxd3 d6 7.
>>>> Rfe1 Nc6 8. Qe4
>>>> 13 2:03 0.28 1. ... cxd4 2. cxd4 Nc6 3. O-O d6 4.
>>>> a3 Rc8 5. Rac1 Nd5 6. Bxe7 Qxe7 7.
>>>> b4 f5 8. b5 Na5
>>>> 13-> 2:46 0.28 1. ... cxd4 2. cxd4 Nc6 3. O-O d6 4.
>>>> a3 Rc8 5. Rac1 Nd5 6. Bxe7 Qxe7 7.
>>>> b4 f5 8. b5 Na5 (s=2)
>>>>
>>>>IE until 14 secs it would not play h6, then it likes it until 42 seconds, and
>>>>then it won't come back to it in any reasonable time limit on my xeon...
>>>>
>>>>Given enough time it might actually change to d5, as it looked at that move
>>>>longer and longer every iteration...
>>>
>>>That's strange. All my Windows Crafty engines (19.06,19.08,19.09) like
>>>...h6 or ...cxd4 2.cxd4 h6 at depth 12 and 13.
>>>I don't have a Linux Crafty to test.
>>>
>>>Nd5 is good. The position looks equal after that move.
>>>
>>>Michael
>>
>>
>>The 19.9's are a bit "iffy" as I have made many changes in cleaning things
>>up. And I wouldn't be surprised if one or two of the changes affected this
>>as code was cleaned up here and there.
>>
>>Also are you playing in game mode or analyze (analysis) mode? That makes
>>a huge difference in what it thinks about king safety. IE analysis mode
>>turns off almost all asymmetry that can be turned off, which will certainly
>>change moves a lot...
>
>Please pardon my ignorance, but what is "asymmetry," why do you have/use it, and
>what does it buy you? Does it help a lot?
>
>Bob D.
It addresses a potential problem. If both sides castle king-side, and Crafty
is playing white, if it opens the g-file, it hurts its own king safety, but it
also hurts the opponent's king safety. That is a perfect whitewash. Asymmetry
means that the evaluation is asymmetrical. IE crafty values its own king
safety more than the opponent's king safety, so it won't shred its own king's
position to expose the opponent's. This is a "conservative" approach that has
drawbacks, but it also solves problems that happen more frequently in fast
games like blitz.
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.