Author: leonid
Date: 06:49:43 01/20/01
Go up one level in this thread
On January 20, 2001 at 05:41:48, Uri Blass wrote: >The competition should not be on playing chess but on being faster. > >The target should not be to win games but to calculate the sum of the evaluation >functions of all the possible positions 7 plies forward(it means that the >programs do not need to use the alpha beta but to do a full search of all the >possible positions). > >A possible evaluation may be for example the following material +mobility >evaluation: > >material evaluation:pawn 100,knight 300,bishop 325,rook 500,queen 900 > >mobility evaluation: > >bishops and knights get evaluation for controlling squares by the following >rules(there is no logic behind the rules): > >Bishops or kings: >1 -2 3 4 3 2 1 2 >2 3 4 5 4 5 3 2 >3 3 3 4 5 3 2 6 >9 1 2 4 6 3 2 1 >0 2 4 3 2 1 3 4 >6 4 2 1 3 4 5 6 >1 1 1 1 0 4 3 2 >0 2 0 3 2 4 2 1 > >Knights,Rooks or queens: > >1 -2 12 12 13 12 0 0 >0 0 0 0 0 0 0 0 >0 0 0 0 0 0 0 0 >1 2 3 4 1 2 1 0 >1 -2 12 12 13 12 0 0 >0 0 0 0 0 0 0 0 >0 0 0 0 0 0 0 0 >1 2 3 4 1 2 1 0 > > > >The first line in the bishops or king matrix (1 2 3 4 3 2 1 2) means that >every white bishop get positional bonus of 0.01 pawn >for controlling a8,0.02 pawn for controlling b8...0.01 pawn for controlling g8 >and 0.02 pawn for controlling h8. > >For the black pieces it is the same when we talk about controlling a1-h1 instead >of a8-h8. > >The kings also get the same bonus for controlling the same squares. > > >The target of the programmer is to do a 7 plies full search from the initial >position and to calculate the sum of the evaluations of all the legal positions. > >Hash tables may help because the program can use them not to search positions >after 1.e3 e6 2.d3 when it searched positions after 1.d3 e6 2.e3 and to remember >the number 2 for 1.d3 e6 2.e3(this position can be achieved in exactly 2 ways >after 3 plies). > >It is possible to use a random position for the competition in order to prevent >programmers to tune for the initial position. > >The winner will be the program that is the fastest in calculate the sum of the >evaluations of all the legal positions. > >I think that it may be possible to learn from the winning program about ways to >improve the speed of chess programs. > >It may be also possible to learn about better way for writing an evaluation >function or better ways of representing the board. > >The competition can be in 2 categories: > >1)C category (the programmer has to use only C for his(her) program) > >2)General category when the programmer can use everything for his(her) program. > >A lot of people develop free chess programs when they earn practically nothing >for it so the only question is if there are people who want to participate in >this kind of competition. > >Uri I would like to see something beyond what you propose. Some kind of second competition (your is OK), more as collaboration and for different reason. First - give to the new programmers possibility to debug their basic part easely. Until now I never saw such a tool. This will be possible if each program will use the same value for each piece. Just like you said. When given position will be searched at indicated depth. When for each root move final value will be indicated after the search. To make comparison perfect, nothing beyond logic must be used. Brute force search with no extensions. No look up table... Until now only logic that search for mate was 100% verifiable. Existe many good Web Sites to help in this. Give to the new programmers some initial idea what the raw speed of actual program represent. How important this raw speed really are? For making this possible, each presented program must use anything that make comparison obscure. Hash table, selective search and so like. Only brute force. This could give, at least, some initial indication what based raw speed must existe before program goes into its finalization. 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.