Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Nunn test is good!!

Author: José Carlos

Date: 09:41:12 10/07/99

Go up one level in this thread


On October 07, 1999 at 10:05:50, Robert Hyatt wrote:

>On October 07, 1999 at 01:24:47, Jouni Uski wrote:
>
>>On October 06, 1999 at 09:44:36, Robert Hyatt wrote:
>>
>>>On October 06, 1999 at 07:09:18, Jouni Uski wrote:
>>>
>>>>On October 06, 1999 at 05:39:01, Georg v. Zimmermann wrote:
>>>>
>>>>>20 games far too few to make any judgement. Did you delete the learning files ?
>>>>>Change yout time control a little bit, and see what happenes. You'll be
>>>>>suprised.
>>>>>
>>>>>Regards,
>>>>>
>>>>>--Tec
>>>>
>>>>My point just is, that 20 games is enough! Even if statistics say You need
>>>>100s of games...
>>>>
>>>>I don't enable learning ever. If I change time control a little bit, no change
>>>>in scores.
>>>>
>>>>Jouni
>>>
>>>
>>>Why play a match with learning off?
>>
>>Because under Fritz learning changes Fritz general book. I don't want all
>>programs to modify it. Specially in blitz matches it makes no sense.
>>
>>Jouni
>
>
>It does make sense in Fritz.  They use the same learning algorithm I use in
>Crafty, which means they 'scale' the learning results based on the search depth.
>Nothing wrong with learning from blitz games, just so you can tell the
>difference between blitz and standard somehow.
>
>However, playing with 'fritz book' doesn't make a lot of sense, if a program
>has its own book.  Why force them to follow lines that someone thinks is good
>for fritz?

  Anyway, Bob, how does Crafty solve the problem that Fritz resets the game
every move, forcing to clear hash tables. Another problem is that with log file
activated, a game under Fritz generates so many log files as moves the game has.

  José C.



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.