Author: blass uri
Date: 08:36:04 06/23/00
Go up one level in this thread
On June 23, 2000 at 11:13:04, leonid wrote: >On June 23, 2000 at 10:25:17, blass uri wrote: > >>On June 23, 2000 at 10:07:15, leonid wrote: >> >>>On June 22, 2000 at 07:30:50, blass uri wrote: >>> >>>>On June 22, 2000 at 06:27:09, leonid wrote: >>>> >>>>>On June 22, 2000 at 02:54:43, blass uri wrote: >>>>> >>>>>>On June 21, 2000 at 21:18:07, leonid wrote: >>>>>> >>>>>>>On June 21, 2000 at 19:03:40, Tom Kerrigan wrote: >>>>>>> >>>>>>>>On June 21, 2000 at 17:07:06, leonid wrote: >>>>>>>> >>>>>>>>>On June 21, 2000 at 13:38:41, Tom Kerrigan wrote: >>>>>>>>> >>>>>>>>>>If you think that material-only evaluation programs are good for anything, >>>>>>>>>>you're sadly mistaken. >>>>>>>>> >>>>>>>>> >>>>>>>>>I said only that material evaluation is evaluation about everything in principe. >>>>>>>>>About tactics... or just say it. I agree that in actual state of hardware it is >>>>>>>>>not enough have only material evaluation, but its importance will grow as >>>>>>>>>rapidly as hardware capacity will improve. Very soon program that have in its >>>>>>>> >>>>>>>>Only evaluating material has zero importance. Why would you do it when you can >>>>>>>>evaluate material AND positional terms with no penalty? Besides, material is >>>>>>>>just a rule of thumb, just like any positional term. Thinking that you can make >>>>>>>>a good program by only considering material is absurd, no matter how fast your >>>>>>>>computer is. >>>>>>>>-Tom >>>>>>> >>>>>>>Do we speak about my program or about general idea? If we speak about my program >>>>>>>it is not that interesting, since we will talk only about one program in >>>>>>>particular. When we speak about general idea, yes, material echange can say >>>>>>>everything. Only through the material echange you can find mate or draw. By the >>>>>>>same mean you can find all other move in the game, name it positional, tactical >>>>>>>or otherwise. We can talk how much computer power we need for the best program >>>>>>>right now to find this or other kind of move, but this is something else. Idea >>>>>>>is simple - material echange do everything and everywhere. In chess game logic >>>>>>>is enough to see everything in it from beginning up to the end. >>>>>>> >>>>>>>Leonid. >>>>>> >>>>>>Thoretically you are right but practically >>>>> >>>>>So, we say the same. >>>>> >>>>>>Tom is right that material only is absurd >>>>> >>>>>Here it is only the game of the words but actually we are saying the same. >>>>> >>>>>>You do not need material but you need only the 32 piece tablebases. >>>>>> >>>>>>It is theoretically possible. >>>>> >>>>>>If the computer dimensions are 1000,000 kilometers*1000000 kilometers*1000000 >>>>>>kilometers and if it can remember one position in 1/10000 milimeter*1/10000 >>>>>>milimeter*1/10000 milmeter then it can remember 10^48 positions >>>>>>and I know that it is not bigger than the number of legal positions in chess >>>>>> >>>>>>Of course this idea is absurd like the idea of material only evaluation. >>>>>> >>>>>>Uri >>>>> >>>>>Ura, when I tryed to write my first logic for solving the mate I was curious for >>>>>how long ahead people can see (and rapidly) when the mate is there. I found that >>>>>actually it is not that far away, only some 6 or 8 plies deep. Biggest part of >>>>>all "genious, "incredible", "magnificent" move, found by the best champion of >>>>>the world, in real game, during the chapionship were very specifique. Almost all >>>>>of them was instantly solvable by so called "quick mate solving logic" and was >>>>>in the depth between 10 and 14 plies. If human can see actually all moves in the >>>>>game and rapidly, beyond mate and draw, at the same depth as it is for mate, we >>>>>are close to be there. Very soon brute force search for material echange (no >>>>>extensions) will be able to go easely 8 plies deep in around 1 second. This >>>>>could permit to search pretty well by quick logic 14 plies deep to make good >>>>>move. The rest in the game could be easely available by using the database for >>>>>beginning and the end of the game. The extras will be more for overkill that by >>>>>making the program strong. >>>>> >>>>>Leonid. >>>> >>>>I believe that player with rating 2000 will have no problem to win against only >>>>material evaluation,no extensions,14 plies+opening book. >>>> >>>>I believe that 8 plies of TSCP are worth more than 14 plies of only material >>>>evaluation program. >>>> >>>>It is easy to get programs out of the opening book in a few moves so it is not >>>>going to help much. >>>> >>>>Even without going out of book it will be easy to win the 14 ply program(for >>>>example the 14 ply program will not know that it should push the pawn forwards >>>>and it may do stupid mistakes in the endgame by playing passively). >>>> >>>>Tablebases also are not going to help because the program is going to have no >>>>chance before the very simple endgame. >>>> >>>>Uri >>> >>>When you play quick game you hardly will have that much time to thing about >>>everything. Quick game, that so many people like, is mainly the place where >>>chess program is better that normal human and where "brute force" is so >>>important. >> >>Quick game against humans are not interesting because humans lose against top >>programs of today. >> >>I think that only material evaluation will have problems even in quick games >>against humans(not against most players but certainly agaisnt grandmasters. > >Maybe. > > >>In quick game between computers evaluation is more important and I am sure that >>14 ply brute force with only material evaluation is going to have big problems >>against 12 plies+some knowledge like the knowledge of TSCP. > >Don't know what is TSCP. TSCP is Tom Kerrigen's simple chess program. It was written in order to help programmers and not in order to win games. It does not use hash tables and does only some simple extensions. It has a simple evaluation function but the evaluation function is clearly better than material only evaluation. I am not sure what is the 12 plies+some knowledge. But >Probably this "some knowledge" is nothing more that search after 12 plies, >nobody knows how many next plies ahead, in order to recognize prone to drastic >change positions. Positions where some check accur, or some material echange >take place. In reality, all this 12 plies search (if this is exactly what I >think) is a fake. It is the same fake like "brute force", "fixed depth" search >in chess game. Instead of indicated 12 plies it can go 14 or even more plies >deeper. I believe that your program(fixed depth 8) is going to lose against modified TSCP with no extensions at depth 6 and it is easy to check it. TSCP source code is not a secret so you can modify TSCP to do no extensions and test your program against TSCP in order to see how many plies you lose by only material evaluation relative to a simple evaluation. I believe that the difference becomes bigger when the depth of the search is bigger. This additional search is called "extensions". So, in reality we don't >speak about 12 plies search but search of unknown depth. This kind of search >with extensions make me think only about "quick logic" in my program. > >> I don't speak about game where computer think one second and human 5 >>>minutes. Also 14 plies search permit to the program to see almost everything >>>that will happened with the pawn, even when it will reach opposite side of the >>>chess board. Pawn will know, for instance, without any artificial incentive, >>>that he must go ahead and reach its promotion many moves later. Many anomaly in >>>pawns structure (like doubling of the pawns) will be avoid because program will >>>see far enough to see some material loss related to it. >> >>There will be some cases that programs will see that the pawn should go forward >>because of search but in most of the cases they will not see it. > >For this the next few plies will give the next force. And this will go until >force of chess program will leave human in the position of helpless inferiority. >At least, for now it is not that close to expect that Chess Club will become >empty. > > >>It will have more chances if you reduce the dimension of the board to 6*6 and >>increase the number of plies because you can search deeper in a small board. > >Pretty interesting. It never came to my mind. And I am not sure it this is >practical. It could work only if we can expect that few squares will be never >demanded for occupation. The LOS Alamos program from 1956 played 6*6 game with similiar rules to chess when there were no bishops on the board. It could win a beginner in this game but lost against a strong player when the strong player played without a queen. 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.