Author: leonid
Date: 04:43:27 10/04/99
Go up one level in this thread
On October 04, 1999 at 03:06:20, Jeremiah Penery wrote: >On October 03, 1999 at 16:31:38, leonid wrote: > >>On October 03, 1999 at 15:07:58, Ed Schröder wrote: >> >>>>Posted by leonid on October 03, 1999 at 08:26:51: >>> >>>>I am not sure what you want to say with "extention" but my trying was very >>>>simple. The identical positions were tryed on Rebel and Hiacrs using same >>>>computer. Number of plys demanded were the same and from boths games it was >>>>asked to make search by "brute force". Rebel was incredebly speedy. How >>>>raw speed is implemented in final game it is the next question to ask. But >>>>potential of the game is on its raw speed anyway. >>>> >>>>To Rebel was said: >>>>1) Brute force - on. >>>>2) Hash table - put to zero. >>>> >>>>To Hiacrs was said: >>>>1) Selectivity - put to zero. >>>>2) Hash table - put to lowest value possible, to 1024k. >>>> >>>>So, it is true that Hiarcs did his search at brute force and he really a >>>>looser? Great mystery for me. >>>> >>>>Thanks to everybody who will make this enigma vanish! >>>>Leonid. >>> >>>The keyword to understand the topic = extensions. >>> >>>Chess programs do use all kind of extensions and all chess programs do >>>it another way. Some programs do not emphasize on extensions other >>>programs heavily use extensions. >>> >>>A common example of an extension is to extend the ply-depth with "1" in >>>case the king is in check. Thus instead of searching 6 ply the remaining >>>tree is searched for 7 plies. If there is another king-in-check case found >>>the search is extended to 8 plies and so on. >>> >>>So you see a 6-ply search isn't a 6-ply search as a lot of variations in the >>>tree are searched deeper. >>> >>>Some program will extend king-in-check cases even more. On some types >>>of checking moves Rebel will extend 2 plies. Other programs have their >>>own ways of doing things. The end result is that you can't compare programs >>>based on ply-depth. >>> >>>Ed >> >>This explanation puzzle me. My impression was that "brute force" in given >>number of plys deep should include nothing but number of plys asked. Every >>other extra ply for every other extra reason is inappropriate. I can only >>understand that in all games logic should go some extra plys (or extention) >>if it like to do this but not when "brute force" is asked. To see >>some extra ply in "brute force" search (even partially) is pious fraud. It sound >>like find the mate in 5 moves but announce it as mate in four. >>Brute force search in certain number of plys signify for me search for all >>the possiblities that existe for given position but going only the number >>of plys said to be performed. If "brute force" is what I see it to be, it says >>more that every other aspect of the game about the general capability of the >>chess game itself. >> >>Leonid. > >What you are describing is a pure minimax search. No program does this. If a >program tries to search _every_ possibile move to a given depth, it will take a >very long time for all but the lowest depths. It would be pointless. All (?) >engines use alpha-beta of some kind, which does not allow searching of _every_ >node in the tree. >I've read in other posts that you're worried about the branching-factor. If you >do this kind of search (pure minimax), EVERY program will have the SAME >branching-factor. > >Also, I'm puzzled. Why are you so worried about the "brute-force" search? For >writing a chess program, it has almost no relevance. What exactly are you >trying to determine? If you could narrow it down a bit, perhaps someone would >be able to help you more. :) > >Jeremiah Probably my wording was wrong somewhere. For me "brute force" signify the search of every possibility in each ply but not every position. So even for alpha-beta search brute force presumably should give the idea at what speed game do its work and its efficency as well. I need to know my speed in my "positional logic" before going into finalization of my game. The ending of the game is writing the database for beginning and the end of the game. But what sense have for me to finalize the game that is anyway very weak? I must make its engin as strong as it should be. One part of my game was already finished few years ago. I stopped working on it when I found that it run better that every other best game that I could find in that time. It was logic for solving the mate. But compare its speed was very simple. Just take few hundreds positions with inevitable mate and see he do the work in shortest time without any mistake. But how about positional move? It is still a mysery to know where I am. Must I stop there, or I must speed this part even farther? Question that is so difficult to respond. Now, probably, you see what I am been looking for. Leonid
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.