Author: Omid David Tabibi
Date: 13:13:59 12/18/02
Go up one level in this thread
On December 18, 2002 at 16:02:38, Bruce Moreland 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. > >If I were testing the properties of a specific isotope of a specific element, I >would assume that previous research is valid, because the test material is >invariant. > >If I were testing something whose properties were variant, I would have to >repeat some previous research. > >If you understand a certain type of wood, and are skilled at doing fine >carpentry with this kind of wood, you may not be able to use the same methods on >a different type of wood, because that wood may have different properties. > >The blind assumption that all wood is the same would lead you to produce some >crusty looking furniture. > >You can't make use of much previous research in the computer chess field. A lot >of it was conducted on slow hardware, a lot was conducted before null-move >happened, etc. > >So if someone says that R=2 is better than R=3, there is *no way* that I am >going to believe this until I run it myself. But when you run it yourself (what I did) and see that std R=2 is better than std R=3, will you publish it? No, because it's a known result. But if you run it and find that std R=3 is better than std R=2, only then you may publish it, for it is in contrary to the previously published research. > >There are plenty of techniques that haven't been repeatable. I believe that >people have had a hard time repeating MTD(f). >There were also problems >repeating Donninger's "deep search" aspect of his original null-move article. >I'm sure that there are others. MTD(f) worked for Plaat, "deep search" worked for Donninger, Adaptive Null-Move Pruning and Extended Futility Pruning worked for Heinz, Verified Null-Move Pruning worked for me, etc. You have to implement all these ideas to gauge their performance in your programs. > >Most of that research is just ideas about stuff to try to repeat. > >>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 > >I'm assuming you mean R=2 in the second line. Your own data implies this. I >think it behooves you to investigate it. > >I believe that if you run R=3 for the amount of time that it takes you to get to >depth=10 with R=2, *you* will find that you get more answers on both the WCS >suite and the Neishtadt suite. > I conducted self-play matches between std R=2 and std R=3. The results showed that std R=2 is superior, and that was enough for me. >If you don't, I would have serious questions as to why not. You are within one >solution already. How could you expect that you won't get at least two more >solutions if you more than double the time? > >bruce > >>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.
This page took 0.01 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.