Author: Tord Romstad
Date: 09:07:05 12/15/00
Go up one level in this thread
On December 14, 2000 at 12:05:29, Ian Osgood wrote: >On December 14, 2000 at 10:36:08, Roy Eassa wrote: > >>If Genius 1.2 on the Palm is the best by far, yet it's based on a 1987 program, >>I would think that a good Palm programmer (I know somebody who has apps on the >>shelves of CompUSA) and an existing top-level PC chess programmer could >>collaborate to make one that's quite a bit stronger. >> >>Would there be an adequate market for this? > >I wouldn't bet that you would come up with anything better... > >The Palm puts many limits on program strength. > >* tiny heap, so no hash tables >* small storage, so no endgame databases and small opening books. >* slow processor > >Richard Lang's Roma program was his last that didn't use hash tables, which >means it is the acme of what he could produce for this hardware. His program is > written in 68k assembly, which is another reason why it is so strong. And it >embodies his many years of experience in programming slow dedicated computers >(remember, at the time Roma was released, Richard's programs were dominating >computer chess in the Mephisto line of computers). > >The only things you could improve upon might be recent advances in search >techniques (such as null move). I am fairly sure the null move heuristic was used before 1987, although it was of course not as widely known as it is today. Anyways, Richard Lang never seem to have introduced null moves in his programs, so I assume that whatever selective search techniques he uses work better (at least in his programs, and on slow hardware). I personally believe that Lang's search algorithms are still the best on slow hardware, and that they are not easy to improve upon. >Unfortunately, most recent advances presume >hash tables and deep searches (i.e. memory and processor speed to burn). True, but Lang's programs never used much memory (except for hash tables, of course). The executables of the DOS versions of Chess Genius were ridiculously small. With small hash tables, Genius 5 works fine with the 640k memory limit in DOS. I find it very hard to believe that none of the many improvements Lang has made since Mephisto Roma can be implemented successfully on the Palm platform. Hopefully he will work to improve the strength of the program in the future (from what I have read, it seems that the engine in versions 1.1 and 1.2 has not changed noticably compared to version 1.0). > Also, >a root evaluation strategy (used by Morsch's Fritz and Kittinger's Novag >Sapphire/Diamond) could also be successful on slow processors, since it >increases raw search speed. It seems to me that Lang's programs are preprocessors, too. There might be some leaf evaluation as well, but I am quite sure most of the evaluation is done at the root and is updated incrementally as moves are made and unmade during the search. >That said... if you came up with a Palm program stronger than Genius, I'd buy it >in a heartbeat. :) I am trying to find good arguments for buying a Palm. I would really like to run Chess Genius for Palm, but I find the machines a bit too expensive when the main reason for purchase is running a chess program. Unfortunately I am not able to see how owning a Palm could be useful for me ... Tord
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.