Author: Thorsten Czub
Date: 04:51:30 09/10/02
Go up one level in this thread
On September 10, 2002 at 06:23:12, Joachim Rang wrote: >Dear Thorsten, > >I think you missded the point: Martin was askingyou, if you have in mind, that >you need a lot of testing to make significant results. he is wrong (or you too). you don't need a lot of games to prove your style is right. why ? because you watch the games. WHEN you believe you have a good style, you can begin with testing quantity, and if your thesis was right, you will find evidence in the results. so: it can be measures by the success (results) the style gets. if it would not work this way, you could not produce good results. >Otherwise you could be >misleaded by a few games and change the settings in the wrong direction. if you are changing the style in the WRONG direction, you will get losses. so the success of your method can be measured. same with decisions which version against which program or which opening against which program on championships. the success shows if you'r decision was right. if you choose the wrong version, you will lose the game. > I'm >sure you know how to do it right. But in the end, if you have found a probably >better setting, there is no other way to prove it, than to play lots of games >(100 and more) against different opponents to get statistically significant >data. exactly. but during the development process, it is done different than he /you suggests. it is NOT the autoplayer that makes the changes "right". it is the human beeing. when jeroen noomen or alex kure choose wrong opening for their game, they will maybe lose it. so their decision is overall represented by the results they get.
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.