Author: Ed Schröder
Date: 12:24:28 04/06/01
Go up one level in this thread
On April 06, 2001 at 14:09:09, Robert Hyatt wrote: >On April 06, 2001 at 12:34:57, Thorsten Czub wrote: > >>On April 06, 2001 at 11:49:27, Ulrich Tuerke wrote: >> >>>Since nobody is posting wins of my program, I mayself have to do it. -:) >>> >>>Having got curious on Gambit Tiger, I have made 2 consecutive games on FICS, >>>comet even running on a bit inferior hardware (700 MHZ - 1 GHz, time control is >>>4+6.) >>>The games below demonstrate the high risk of gambit tiger's playing style. I can >>>hardly imagine a top engine giving this good chances to a middle class freeware >>>program. >>> >>>The games ended 1.5 - 0.5 for comet, where tiger rather luckily escaped into a >>>drawn ending in 2nd game. >>> >>>These 2 games do of course not imply that comet is stronger. I'm sure, that the >>>ship will sink in a multi-game match. However, they do demonstrate that far >>>weaker opponents have a unexpectedly good chance against due to tiger's risky >>>playing style. >>> >>>I personally doubt that the gambit version of the program can do better than the >>>usual tiger against computer opponents. >>> >>>Regards, Uli >> >>i cannot speak for gambit-tiger2 so far, >>but i can guaranty you that for gambit-tiger1 in relation to rebel-tiger13 >>the success is on gambit-tigers side. >> >>i think the explanation is easy ! >> >>gambit-tiger will lose a few games due to risky style. >>but it will not lose as much games due to doing nothing as >>rebel-tiger13 will do. >> >>if you have 2 similar programs, and the one is more agressive, >>the more aggressive will IMO make a better score over all. > > >I totally disagree. I will be happy to give you an immensely aggressive >version of Crafty. But I believe it will lose _way_ more than it wins when >compared to the normal version. There is aggression, and there is "aggression". >and too much of _anything_ is not a good thing. Exactly! But the BIG difference is that Christophe has managed to make a version (Gambit Tiger) that is EQUAL in playing strength and that is precisely the special thing. You can set Rebel to "Bluff Chess" or create a special personality and it will behave similar to Gambit Tiger. BUT... its playing strength is lowering and the playing strength of Gambit Tiger is not. That is why "Gambit Tiger" really deserves its "Gambit" name. Ed >>if you have a strong chess program, that is not trying to get the initiative >>or is doing almost nothing than waiting for opponent mistakes, >>it will not be very strong. > > >If a program tries _too_ hard to be aggressive, the passive program can safely >sit back and wait for the aggressive version to fall apart. At that point, the >only question is "What has the aggressive program given up to this point?" If >it is a positional weakness, it might not be critical. If it was a piece for >a pawn and dangerous attack, and the attack doesn't work, it will most likely >lose. > > > >> >>the reason fritz and junior do get high scores in schweden is not because these >>programs DO something. but because of book-learning effects and >>outsearching the opponent. >>but if they would have to play against itself, and ONE engine would be >>a little more DOING something, the one that tries to get the initiative >>would IMO win. > > >Initiative is one thing. uncontrolled aggression is something else. I haven't >broached the tiger aggression issue in a good while, because it didn't lead >anywhere. But Uli has a point. When I used to play in our local chess club >a lot, I found myself able to beat 1700 players almost at will. Unless I >decided that "I am going to attack and I don't care what he does to try to stop >it..." When I played like that, I found myself losing more than I expected. > >Aggression is ok, but too much is very likely worse than not enough. >Particularly if you can't out-search your opponent which lets him "call your >bluff and stuff it."
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.