Author: James T. Walker
Date: 11:39:56 03/21/02
Go up one level in this thread
On March 21, 2002 at 13:47:25, Terry Ripple wrote: >On March 21, 2002 at 06:49:08, James T. Walker wrote: > >>On March 20, 2002 at 21:03:04, eric guttenberg wrote: >> >>>Jim >>> >>>I don't know which is stronger but I agree wholeheartedly with your >>>statement that the only way to test is with many games on 2 equal >>>computers. A lot of people (like myself) simply don't have 2 computers >>>and playing matches on a single machine is the only option. It can be fun >>>and can be useful for some purposes but measuring the relative strengths >>>of various chess programs should be done in the environment for which they >>>were designed. >>>For that reason I have long felt that the results you publish are among the >>>the more probative ones seen on this site. Thank you. >>> >>>eric >> >>Hello Eric, >>Thank you for your kind words. Just to gather some more data on F7 & CT14 I >>played some more Blitz games last nite. To my surprise Fritz 7 beat Chess Tiger >>14.0 very nicely by 38.5-23.5. This is the first time either one has won by >>this wide margin. But what I did was to "load up" Fritz 7. Meaning I cleared >>the Fritz 7 book and then let it learn from all my databases. This gave Fritz >>the "experience" of about 1700 games. This seems to be a clear advantage for >>Fritz since I cannot do this with Tiger in the Chesspartner GUI. I haven't had >>time to inspect the games yet so I'm still suspicious that something else may >>have been wrong. If not this is quite an eye opener. >>Jim >------------- >Hello Jim, > >This feature of Fritz 7 that allows it to learn from all your databases, is this >similar in comparison to Rebel Century which if allowed to use the Big ECO will >then be playing from this database. I like this feature of Rebel Century because >it first evaluates the move it's considering to use from it's database to make >sure it's a good move and not a blunder before it actually uses the move! > >Regards, > Terry Hello Terry, It is very different from Century and ECO. This is only book learning with Fritz/Chessbase GUI. I'm still not sure what happened and if the score was caused by the learning feature. In Century and ECO, Century will take the position from ECO and consider the move first. But if not set up in the "BOOK" mode it only uses this move for first choice. It can change it's mind with the search. This is a feature which I like but has the drawback of learning the wrong move in positions during a game. It may then save the position and later use the wrong move for "First choice". Fritz just learns which openings it's winning with and which ones it's losing with I think. I still have some testing to do to determine what caused the big jump in score by Fritz. There are many variables here because I'm using 2 computers. Each already has it's own learning but doing the 1700 games seems to have really upset the applecart. I have switched computers and now CT14 is beating Fritz 7 by 25-15 after 40 games. I see nothing wrong/different about the programs NPS so I don't suspect a problem with the machines. Back to testing. :-) Jim
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.