Author: Uri Blass
Date: 15:21:33 10/03/03
Go up one level in this thread
On October 03, 2003 at 17:34:56, Christophe Theron wrote: >On October 03, 2003 at 15:36:22, Mridul Muralidharan wrote: > >> >>You misinterpreted me. >> >>On October 03, 2003 at 14:51:24, Uri Blass wrote: >> >>>On October 03, 2003 at 13:38:54, Mridul Muralidharan wrote: >>> >>>>Hi, >>>> >>>> I was a bit taken aback by these declarations : >>>> >>>>On October 03, 2003 at 12:47:23, Uri Blass wrote: >>>>>I prefer even not to care about using hash tables for pruning because my >>>>>experience told me that I cannot get significant gain there easily >>>> >>>>Hash table not giving you pruning ? I suspect a bug in your hashkey - >>>>nothingelse. >>>>Or maybe it is the easily that is operative word ? >>>>I think there are a lot of open source programs that you can refer to and >>>>correct your bugs with - crafty , GNUChess , etc , etc. >>>>Might help to get this right. >>> >>>I do not like to copy from other sources. >>>I found that instability helped me to do my program significantly better. >>> >>>If I delete it in order to be able to copy from other programs then I may need >>>to start by doing it significantly weaker. >>> >> >>I did not mean - "copy" here. >>Rome was not built in a day. What I meant is : >>Look at their implementation - check yours. Find any obvious bugs. >>I seriously suspect that there are - since hashtables not only help in pruning , >>but massively help in move ordering. >>If you can afford to make these statements - then your impl is horribly full of >>bugs. >> >>As far as "instability helping" - I'm really not sure what you mean by this. As >>far as I know - everyone , including me , tries to reduce instability so that >>search is more stable requiring minimal search tree. >>Wild extensions , unstable pruning , etc may help you in solving test suites >>better and faster - but in real world games , it will suck badly. >> >> >>>> >>>> >>>>>(I have a lot >>>>>of stuff that means that pruning or extension is not defined only by the >>>>>position). >>>> >>>>Where ever possible , I try to make the search behaviour as relevent to the >>>>current position as possible and not rely on past search. >>>>Why do you want to do the opposite ? >>> >>>because the opposite gives me some advantages. >> >>test , test , test - dont assume. >>like my collegue says : When you AssUMe , you make an Ass of U and Me ;) >> >>>Movei has its chances against every program inspite of having bad order of moves >>>and bad extensions and bad pruning. >>> >> >>acceptance is the first step to improvement ! >> >>>I believe that I can get above Crafty level if I improve order of move >>>extensions,pruning and evaluation. >>> >>>Movei already has its chances against Crafty but today crafty is significantly >>>better. >>> >>>There is a lot to improve and the main problem is programming. >>> >>>Uri >> >> >>AFAIK movei is not smp - so no point in saying search here :) >>SO , other than move ordering , eval and pruning : what else is left ? interface >>code ? ;) >>anyone can get to crafty level or higher - IF you are willing to put in the >>effort and scientifically research. >>All the best - wishing to see a better Movei and a more scientific Uri :) > > > >You are talking without knowing Uri, obviously. > >You are also talking about commputer chess without really knowing, I fear. > >Uri has been talking with us about chess programs for a long time without >writing one. So at the begining he came with ideas that were not really >relevant, or not exactly to the point. > >What I like about Uri is that at some point instead of keeping talking about the >subject he has started to write his own chess program. I also know that he has a >very scientific, objective approach in what he does. > >So if I was you I would not give Uri beginners's advices like "test, test - >don't assume". Because as far as I know that's exactly what he is doing. > >I also believe that Uri is right in not trying to get too much inspiration from >other chess programs. > >I think you should refrain from giving lame advices to Uri and tell us instead >about your achievements as a chess programmer. > > > > Christophe Thanks for the support. I believe that I did not implement most of the ideas that I had before writing a chess program. There are things that I did not try not because I believe that they are not productive but because they are not simple to implement and I prefer to think about implementing more simple things first. About testing I can say that in a lot of cases I am not sure if the program is better after a change that I accept but at least I am sure that it is not significantly worse. Uri
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.