Author: Ralf Elvsén
Date: 05:33:28 09/01/99
Go up one level in this thread
On September 01, 1999 at 06:35:24, blass uri wrote: >On September 01, 1999 at 05:53:59, Ralf Elvsén wrote: > >>On August 31, 1999 at 22:39:09, eric guttenberg wrote: >> >>>If there is any reliable data showing that CST can even hold its own with >>>H7.32 in a significant number of games, I will be surprised to see it. >>> >>> >>>eric >> >>Could you (or Mark or anyone else...) who have seen CSTal >>play alot, tell me why you think it is weaker than e.g. Hiarcs? >>(I don't own a copy). > >I also don't own a copy >I do not know if tal is weaker than hiarcs but I do not believe that it is more >than 200 elo better than it(Thorsten did not post enough games to prove it but >this is the conjecture that thorsten's results suggest) > >My impression based on games and evaluations that I saw is that tal's evaluation >is not objective and prefers the attacker in unclear positions. > >It does not imply that tal can be improved by using more objective evaluation >because the best evaluation to win games is not the same as the best evaluation >to predict the result. > >>I always thought that it would be interesting to see >>a CSPetrosian, since that style might be better suited >>to the programming technique used in CSTal (not that >>I know anything about it, except for the low NPS). > >1)The assumption that fast searchers are better at tactics is sometimes wrong I didn't assume that. See below. > >2)I think that you always have to avoid tactical errors and I am not sure if >playing in positional style when the games are long can help you. I was thinking like this: CSTal's style gives it some humiliating defeats but also brilliant victories. The last fact shows that it sometimes has the upper hand when it comes to understanding the "strategic" aspects of a position. Now, why couldn't that be applied for some slow positional killing instead of spectacular fireworks? I believe it could. This is under the assumption that it is strong enough tactically, which (as you say) it probably can be. If it has tactical holes it might be possible to add some code that (given the cost is small enough) makes it avoid extremely tactical positions (probably contrary to what it is doing now). I think this would be the best way (although not the comercially most successful way :) to exploit the benefits of this programming approach (whose details I know very little about). The pros are still there: strategy. The cons are gone: actively seeking positions where your (sometimes) "narrow" search proves inferior. I think the question of tactical ability is of minor importance. If it's there now, it will always be there. The problem boils down to this in my opinion, which strategic weakness of your opponent is most fruitful to exploit: King safety or positional ignorance? Ralf > >If you play in tal's style you may discover that the game was so short that you >had no time to do tactical errors. > >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.