Author: Omid David Tabibi
Date: 13:04:55 12/18/02
Go up one level in this thread
On December 18, 2002 at 15:48:50, Bruce Moreland wrote: >On December 18, 2002 at 11:20:58, Gian-Carlo Pascutto wrote: > >>On December 18, 2002 at 11:07:49, Omid David Tabibi wrote: >> >>>Have you ever conducted any research? If so, you would have known that a >>>researcher doesn't examine everything since the creation of earth, he takes >>>something which is known to be better and tries to improve it. >> >>Which is why you investigated Heinz's adaptive nullmove. >> >>Oh, wait... >> >>>I didn't think that someone will seriously claim that std R=3 is better than >std R=3; but now, I'd be glad to write another paper comparing those two, and >>>also mentioning fixed time comparisons if people find it interesting. Because >>>although not appearing the article, I have conducted tens of other types of >>>experiments (including fixed time) and I _know_ that vrfd R=2 is clearly >>>superior to std R=3. >> >>'Everything you know is wrong' >> >>Whether R=2 or R=3 is better depends very much on the search below >>that nullmove. For a Crafty-style program (which Genesis appears to >>be), R=2 is going to be superior over R=3. But you can't claim that >>is a general truth. > >The data in the article strongly implies that for this program, R=3 is better >than R=2. > I didn't publish comparisons between std R=3 and std R=2 because there was no justification for doing so, from a scientific point of view. But that doesn't mean that I haven't conducted self-play matched between std R=2 and std R=3. I few months ago I even posted some of my impressions based on those experiments, and claimed that for longer time controls the superiority of std R=2 over std R=2 is not that significant. Nevertheless, std R=2 always performed better than std R=3. >I think that programs are different enough that if you are going to study >various tweaks of null move, you start by thorougly investigating the vanilla >techniques. > >In computer chess, almost every assumption is baseless, because the programs are >different enough that you'll get different results. > >To hash qsearch or not? Bob doesn't, I do. In mine, it is *definitely* better >to do it. In Bob's, it is *definitely* better not to. Who is right? Both of >us. > >Before I did *anything* with null move, I would first get baselines for all >manner of R. > >I have done tests from R=1 to R=6, with fractional values in between. I don't >do them just once, either, I do them ever year or so, to make sure that things >have not changed. Straight R=2 is best, for me. > >bruce
This page took 0.09 seconds to execute
Last modified: Thu, 07 Jul 11 08:48:38 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.