Author: James Robertson
Date: 13:33:34 02/17/99
Go up one level in this thread
On February 17, 1999 at 15:16:47, Robert Hyatt wrote: >On February 17, 1999 at 14:57:38, James Robertson wrote: > >>On February 17, 1999 at 13:58:38, Vincent Diepeveen wrote: >> >>>On February 17, 1999 at 13:41:33, Will Singleton wrote: >>> >>>> >>>>On February 17, 1999 at 08:18:01, Djordje Vidanovic wrote: >>>> >>>>>Bob, >>>>> >>>>>I read your report carefully. The story is over for me. I know where I am now, >>>>>and what I have been doing. About my doubts -- yes, I did have them but there >>>>>was that different style of play, my good faith and a rationally based >>>>>conviction that the author is a brillant person. For all I know the author IS a >>>>>brilliant person. However, I have enough evidence now to form my opinion about >>>>>those early versions of Voyager -- R=3, some eval changes (e. g. bishops getting >>>>>more bonus vs. knights than you assigned), plus some other Voyager specific >>>>>changes which I will not mention now as the author is still, I think, working on >>>>>the program (hopefully making a wholly new product). Bob, thanks for doing what >>>>>I thought you should surely do -- presenting evidence. From now on, the version >>>>>playing on ICC (a terrific blitz player otherwise) will be labelled as Voyager, >>>>>by Robert Hyatt, modified by G. Mueller. I do intend to run it more when I have >>>>>time, as I truly believe it to be one of the best blitzers on the Net. >>>>> >>>>>There are some other points that I have raised in the discussions with Dan Homan >>>>>and Jeremiah Pennery that I think are worth further elucidation. Perhaps later >>>>>at some point. >>>>> >>>>>Regards, >>>>>Djordje >>>> >>>> >>>>Looking at the results of testing posted on your website, it appears that Mr. >>>>Mueller has done something pretty remarkable; that is, he has found significant >>>>grounds for improvement in a program that has been under development for years >>>>by RH. What, exactly, are the changes that have improved it? For a program >>>>that beats Crafty on equal hardware in *every* single match (as you state), I >>>>think people would be extremely interested to know how he did it. This could be >>>>very exciting. >>>> >>>>Would it be possible to post or otherwise publish the parts of the code that >>>>were changed, so that we may share in this great achievement? >>>> >>>>Re your new labeling of MagusX, etc, I don't know if Bob wrote a program called >>>>Voyager. I thought it was Crafty. Other people have modified versions running, >>>>and to my knowledge they retain the Crafty name. Maybe I'm wrong about that, >>>>don't know. >>> >>>I don't think so that voyager is improved compared to crafty. >>> >>>It uses R=3 and some alpha beta dependant extensions, and a good book. >>> >>>That is: at blitz it seems a little faster because of this R=3, however >>>i'm sure that at a slow match it will not perform better than crafty. >>>It shows *exactly* the same scores and mainlines after say 12 or 13 ply >>>search. >> >>Would it help Crafty to find out how much time it has left, if it is less than >>say, 15 seconds, set R=3, otherwise R=2? >> >>James >> > > >Sounds risky, because you would change the search in the middle of something, >and then would have to compare some scores computed with R=3 to some with >R=2. And they don't compare very well. I was thinking of (in Dann Corbit's words) a function of the time control, something that wouldn't change in a game. James > > > > >>> >>>I doubt whether some alfabeta dependant extensions which solve some >>>problems quicker will *ever* make a program play better in a 3 mins a move >>>match at reasonable hardware. >>> >>>>Will
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.