Author: Robert Hyatt
Date: 09:10:28 03/13/01
Go up one level in this thread
On March 13, 2001 at 11:35:12, Roy Beam wrote: >On March 13, 2001 at 09:39:14, Robert Hyatt wrote: > >>On March 13, 2001 at 00:03:19, ERIQ wrote: >> >>>first I would like to say that I do not understand how the match was not fair. I >>>hear two main complaints: >>> >>>1) The book >>> >>>2) The learning being disabled >>> >>>my answer to 1) is that Yes the book did not favor crafty, but it did not favor >>>CT either as it was CA. 5.1's book. is it good or bad, frankly I'm no expert on >>>the subject but as both programs had to get both sides of the same or simular >>>line, I fail to see how either program gets advantage ! It just proves that one >>>of them could deal with a bad hand better than the other. In short they both had >>>the same problem lines. >> >>This doesn't matter at all. A chess program is a combination of the code, >>the books, and anything else it uses (tablebases, etc) to play the game. In >>the case of your match, there was no variability. If you think playing the >>same games over and over proves anything, feel free to continue. I don't take >>such matches seriously, however, because I realize they are _badly_ flawed. >> >> >> >> >> >>> >>>I noticed also that crafty had good position right out of the opening in several >>>of the games but as time went on and sometimes got short It loss. A human master >>>(fide 2400) would have beaten CT in a few of those games or at worst drawn a >>>few. >>> >>>My answer to 2) how much is the learning feature going to really help in only 10 >>>games anyway, sorry this just sound like a cop out too me. But anyway Crafty >>>reached equal postions in almost all the games. It seemed to go astray more in >>>the late middle game, and in the end game, even in what I thought was winning >>>postions out of the opening. >> >>It will help a significant amount. Otherwise I would never have written the >>learning code. Crafty doesn't like the Sicilian particularly, which is why >>it only rarely plays it on ICC. And as a result, it doesn't particularly >>play it very well. But after playing a number of such games over time, it >>does learn that it does poorly in some of the variations and turns those off. >> >>If you want to play matches with a strange book, feel free. But you should >>also play games with one pondering, the other not. With two different machines >>that are way different in speed, etc. Because the overall result is just as >>bad.. >> >> >> >> >>> >>>Don't get me wrong I do like Crafty I have it as an engine in CA and on the >>>linux side of my box, but this is not the first match I've seen it take a >>>beating from that tiger at 5:00 min game. It just looks like CT is a better >>>blitzer. I've played other matches like 10:00 min game and up and Crafty has won >>>or been even, but at blitz it alway seems to come up short against CT. >> >>Tiger may well be a better blitzer. Or it might be a better blitzer in >>that one opening. Without book learning, you get the same games. That >>seems to be pointless. >> >> >> >>> >>> I think this time Crafty just took a solid beating in ten games nothing more >>>nothing less. But my point in displaying it was just to show that even the old >>>tiger is still very very dangerous. >>> >>>PS. No teeth missing from that cat :) >> >> >>Never thought there were. But there _are_ "teeth" missing from Crafty. Its >>own book and book-learning to name just two. > > > > Why not just admitt that Tiger is the Better Program?? I never said it wasn't. I only said the test was _flawed_. Of course, as always, believe what you want...
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.