Author: Peter Berger
Date: 10:21:26 05/21/03
Go up one level in this thread
On May 21, 2003 at 08:50:15, Uri Blass wrote: >On May 21, 2003 at 08:27:55, Aaron Tay wrote: > >>On May 20, 2003 at 17:33:14, Albert Silver wrote: >> >>>> >>>>Johan spends a great deal of time testing his engine, and his testing results in >>>>the settings that are used in the program. Johan then usually will test out his >>>>theories/suspicions in tournaments such as Leiden. Ultimately, it would be silly >>>>to say that somebody else knows The King better than (or even as well as) Johan. >>> >>>Maybe, maybe not. Ed Schroeder has often sent out contests for users to try to >>>improve on his settings, or find settings with an edge in specific areas, and he >>>certainly tested his program enough. This would seem to show that perhaps he >>>doesn't believe it's impossible for a dedicated user to find an improvement. >>> >>> Albert >> >>but so far no-one managed to find a setting that plays better correct? some were >>better at test suites, but no one managed to find a clearly better personality >>in terms of playing strenght. >> >>perhaps this tells us something? > >That tells us that testers were simply lazy at that time and did not try all >combination of changing one parameter. > >Ed found that reducing the chess knowledge of Rebel from 100 to 25 helps but no >tester was able to find it. > To me this doesn't sound like laziness at all but shows the difference between tests done by the engine author and by beta-testers. It's extremely unlikely that a reasonable tester will change chess knowledge to a quarter of the usual value because it's against intuition. But the major reason is that "chess knowledge" isn't well-defined at all - only the author can have an idea what it is about and what lays behind. On the other hand if you have many testers, _someone_ might come across this and do it by accident. But it's more likely that the author finds out something like this himself. "The King" parameters seem to be defined in a very intuitive way for users btw. I sometimes use changed parameters for analysis if I think that the program over- or underestimates a certain aspect of the position and it usually works just as expected. That's quite amazing IMHO. It might also be the major reason why users were able to find decent improvements over the default settings, although there are probably many internal values that get aggregated to the parameters. 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.