Author: Don Dailey
Date: 13:30:28 09/06/98
Go up one level in this thread
On September 06, 1998 at 15:10:56, Fernando Villegas wrote: >Hi Don: >Glad to talk with you again... if this is talking at all. :-) With respect to >this issue of what makes better..etc.., let me say something very simple from >the point of view of just an user, not a programmer. In my view, after decades >to play computers, what makes a program more efective againts humans is exaclty >the degree they play like computers, such simple like that. >Why? Well, when we say a program play "like computer" I understand we are saying >it play basically on the ground of tactical shots, search, etc., that is to say, >in the area where they are objetively stronger than average player and so, >logically, they should be stronger than us precisely in that area. How you >normally lose against a top program? Simple: losing a pawn after a tactical >combination, sometimes very hidden, sometimes just based in a blunder from us. >If we does not take into account the games where we were goping to lose BUT we >took back the blunder move, the statistic would be more convincing in showing us >we lose there, in the tactical realm. For the same reason we tend to lose sooner >when the program, besides being tactical, is agressive. Fritz is not a paramount >strategist, but it is enough his agressive behaviour and lethal tactics to beat >us very soon. Rebel is more strategically inclined and so he let me some room >to think in peace and I usually get draws against him, although I have never won >a game to Rebel. With Fritz 5,03 -not 5,00- the story is very different. Same >with Junior. Exception made of IM and GM, player lose in any stage of the game >-even in endings- because tactical blunders and/or tacctical shots by the >adversary. And that strenght is based very sustantially, as far as I know, in >searchin ability, in sheer speed, in full and very huge tree searches. >I know that it is supposed that in the are aof tactics almost all has been got >and that now positional play is what is lacking to programs for reaching new >heights; that could be truth, but I feel that a lot must still be done in >tactics and if a programmer concentrate in that fiedl his effort, THAT will be >at the same time the way to do a program better than human in higher proportion >as usual now. >Best regards >fernando I actually believe that if there is an intrasitive relationship between computers and humans (where computer A > computer B but computer B is better than computer A against humans) it will be in favor of the fast searchers being better against humans. I didn't say this in earlier posts but I thought it. I have actually heard both opinions by others so either idea is not new of course. Your reasoning is exactly on the mark as far as I'm concerned, I think the big issue is taking advantage of whatever strengths you possess. So if a computers main strength and advantage against humans is its awesome calculating ability, I cannot see the logic of giving too much of this up for more evaluation. Of course all of this is subject to some simple common sense observations. Your program still needs a lot of knowledge to do well against ANY opponent and this rule should always be honored first. But if your program can get away with trading a little knowledge for speed, I would guess this is at least as good against humans, maybe more so. But it is only a pretty wild guess on my part! - Don
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.