Author: Peter Fendrich
Date: 04:13:13 05/12/05
Go up one level in this thread
On May 12, 2005 at 06:46:43, Rolf Tueschen wrote: >On May 12, 2005 at 01:33:12, Kurt Utzinger wrote: > >>To understand what such a game is worth >>please study Heinz van Kempen's comments >>about Ktulu 7.0 in the CEGT tourney: >>http://www.chessfighters.de/cegt/html/comment_3.html > > >Kurt, just a few remarks on Heinz being surprised through statistics. It is a >very good example of a statistical survey. Under a thousand of games we see a >real yoyo in the results. Is that a proof for uncertainty in this World? I don't >think so. I think that normally people (also testers) simply ignore the Laws of >Stats. Again, if you have so little games, yes, a thousand for thirty programs, >that is little, then you must not be surprised about surprises. > >I want to comment on something more interesting. We could ask the question why >the famous commercial machines don't have such a yoyo in their results. The >answer is easy. Before such a program would be released, it has been tested in >its stability. Sounds easy? yes, but this is the whole trick. Somewhere in the >Sprechstunden (I believe) or in CSS I read something about the development of >professional progs. How the results danced up and down dependig of the >"testpositions" etc. > >Professionals do their tests before and - surprise? - amateurs do their tests >with new updates. ;) I say that because in a real tournament Ktulu wouldn't winn >a point against Shredder. > >Rolf I think you have a point here Rolf. Professionals have a more consistent testing procedure than amateurs and that testing procedure reveales "holes" that they can fix and test over again. Amateur programs often have some strong parts inside but lack of good testing procedure (or time) will make the "holes" to be left there and make the program to play more ad-hoc. Strong in some positions and free fall in others. /Peter
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.