Author: Stephen Ham
Date: 15:35:28 12/30/02
Go up one level in this thread
On December 30, 2002 at 16:49:23, robert flesher wrote: >Without the opening knowledge that we all attain as we become better players the >game is NO longer Chess it a matter of speaking. The computers must have the >basic positions of all the major openings as these are time tested and proven to >be sound in relative terms. I have stated that engines are tuned for certain >positions, this is one of the strength's and should not be taken away from them. >Its like asking, or telling Mikhail Tal he cannot sac.....//Botvinnik would have >thanked us:)//...... there goes his strength, his edge!. If the concern is the >rating?use the Nunn test positions where each engine gets a chance to evaluate a >key, well known position. To watch computer's battle through the opening seems >like a waste of time to me as they still lack the long term strategic >understanding needed. The day i see a computer play the Marshall Gambit, Morra >Gambit, or other genuine pawn sacks in the opening for positional play, or to >get an edge, then i will change my mind. Cheers~ Dear Robert, Thank you for some interesting points. Still, I agree with Uri that there's value in chess engine testing without opening books. In short, I think it all depends upon WHAT one is trying to test. I feel reluctant to post here since I'm a computer dummy, while you and the other readers know more about them than I ever will. As such, I'm at risk for weriting something stupid. But if you can bear with me for a bit, perhaps I can illustrate my point. I first visited this site about the time of my correspondence chess matches with Fritz 6a and Nimzo 7.32. http://www.correspondencechess.com/campbell/ham/ham.htm At that time, a debate was raging about ChessMaster 6000. It was involved in various tournaments with Fritz and Nimzo and Rebel, etc. While never a tournament winner, it generally seemed to finish 2nd or 3rd in each tourney. But after I saw that actual game scores, I concluded that it may be the "strongest" program. Why? Because it was clear that it had the shallowest opening book. While Nimzo's opening book was deep (often extending well over 20 moves) and broad, and the opening books of the other programs were customized for each of their respective styles, ChessMaster seemed to have a very "generic" book that was often out-of-book by move 10. That meant that it had to spend about 20-25% of it's total time to find moves that were still in the opening books of its competition. That's the equivalent of playing at a handicap of 20%-25% less time than your opponent. So, what that meant to me as an observer, was that ChessMaster might be the instrisically strongest program. Instead, it only finished 2nd or 3rd because of it's small and untailored book. So if ChessMaster 6000 had a larger book - one customized to its strengths, it might have been a regular tournament winner. Nonetheless, for those who want an engine with the best computational strength, then ChessMaster seemed to me the ideal candidate. Still, I know that's not what everybody wants. Many here like to have their programs compete in tournaments, and part of that tournament success/failure is connected to having high-quality books. So for those who wish to make comparisons on that basis, then you are absolutely correct, Robert. Then an exclusion of opening books is silly. But for those who instead want to find out which program is the best general analyzer, there's value then in excluding opening books from the testing process. What's my point? Well, some of us chess players want to test out our opening novelties against strong competition that won't tell everyone else about our secret discoveries. A strong chess engine is helpful at such times, and an opening book is of no value to it then. Instead, we are just concerned with accurate analysis. All the best, Stephen
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.