Author: Chessfun
Date: 11:08:19 06/26/01
Go up one level in this thread
On June 26, 2001 at 14:03:14, Bertil Eklund wrote:
>On June 26, 2001 at 09:10:10, Chessfun wrote:
>
>>On June 26, 2001 at 09:02:45, Chris Carson wrote:
>>
>>>On June 26, 2001 at 08:45:11, Chessfun wrote:
>>>
>>>>On June 26, 2001 at 08:12:01, Chris Carson wrote:
>>>>
>>>>>On June 26, 2001 at 07:17:56, Mark Young wrote:
>>>>>
>>>>>>[D]1rb1k1nr/2q1bppp/p1npp3/2p5/1pP1PB2/3P1NPP/PP2NPB1/R2Q1RK1 b
>>>>>>
>>>>>>It is clear looking at the game, IM Matsuura was playing to keep the position
>>>>>>closed against ChessTiger, but ChessTiger refuted this with 17...f5!. Opening
>>>>>>the position and beating IM Matsuura(2467 Elo).
>>>>>
>>>>>Tactics (anti-computer) that work against programs at "blitz" time controls do
>>>>>not work against computers at longer time controls.
>>>>
>>>>That's nonsense.
>>>>The same was said 5 or 6 years ago against Pentium 100's.
>>>>Yet these blitz games you are now looking at are the equal
>>>>of tournament controls on P100's.
>>>>
>>>>>Trojan horse, stonewall, dutch, indian, KIA, KID, ... may be a good strategy
>>>>>(anti-computer), but at 40/2 programs are much stronger than 2100, the results
>>>>>show it, I have tried it with very limited success and the programs adapt
>>>>>quickly to the strategy.
>>>>
>>>>Tell me please of the program that adapted to the strategy and how?.
>>>>
>>>>Sarah.
>>>
>>>Rebel Century, Chess Tiger, Deep Junior, Deep Fritz.
>>>
>>>Try it yourself at 40/2. If you can win 50% in a 10 game match, post the PGN's
>>>and I (and others) will verify. I have tried, but did not work for more than a
>>>game or two.
>>>
>>>I do not know how these programs adapt, but they did using learning or
>>>programming, you will need to ask the programmers, not me, I can only tell you
>>>how my program "Dallas" handles it, but the learning.
>>>
>>>Best Regards,
>>>Chris Carson
>>
>>
>>AFAIK none of the named programs have position learning only book learning.
>>Once a win is found by taking the program out of book, making a few branch
>>factors by analysis, it should be repeatable. Remember for a program to be
>>2600 ELO it should win 10/10 from me.
>>
>>Sarah.
>
>Hi!
>
>Century3 has some kind of position-learning. In fact it is very successfull when
>it is out of book in the same book-line. I have seen it several times in some
>100 game matches against S5 and Gambit. I guess you have seen it, when it lose a
>game it stops and update the learning and checks the auto232.
>
>I thought everyone in the ongoing tournament should try to play the same
>Colle-line in the ongoing tournament as it obviously was out of book in move2.
>
>Bertil
Hi Bertil,
I don't have much luck autoplaying Century, though it seems the SMP
programs are less likely to repeat a move than the single cpu programs. I don't
think though this is a result of programming. Gandalf I think also has position
learning.
Sarah.
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.