 |
|
 Chess variants (10x8)Sam has closed his piano and gone to bed ... now we can talk about the real stuff of life ... love, liberty and games such as Janus, Capablanca Random, Embassy Chess & the odd mention of other 10x8 variants is welcome too For posting: - invitations to games (you can also use the New Game menu or for particular games: Janus; Capablanca Random; or Embassy) - information about upcoming tournaments - disussion of games (please limit this to completed games or discussion on how a game has arrived at a certain position ... speculation on who has an advantage or the benefits of potential moves is not permitted while that particular game is in progress) - links to interesting related sites (non-promotional) Vous n'êtes pas autorisé de poster des messages dans ce forum. Le niveau d'adhésion minimal requis pour poster dans ce forum est Pion. Mode: Tout le monde peut poster
31. Mars 2006, 22:15:41 modifié par Chicago Bulls (31. Mars 2006, 22:17:44) Reinhard: .
. . Personally I think, that huge engame tablebases would not significantly decide the outcome of a game. Of course! You are right. That's what i believe too.... But i'm speaking about 5-6 piece tablebases. But if it is proven that having tablebases makes no harm to the program's strength, then any positive influence due to tablebases is welcome.... Maybe 0.1% of all games could benefit. Hmmmm at Chess i think this 0.001 is a small guess. It must be higher. But at Gothic Chess and CRC games i think it might be correct..... From the Chess games i've seen, i can roughly calculate that in 1 out of 120 games programs have a profit from endgame tablebases.... But to demonstrately use them is an indicator of panic. I don't believe this is the reason! Ed Trice is obsessed with tablebases:-) Remember his huge work on the endgame tablebases for his Checkers program. Now that Checkers' interest is decreasing (the game is "close" enough to be solved- for example to determine the winner between 2 of the best today programs they needed 624 games and the score ended with 4 only wins and 620 draws!!!!!!!!!) Ed couldn't forget his obsession with tablebases and started to generate the Gothic Chess tablebases.... I believe that tablebases hide a brilliant mystery inside them (i'm always amazed when i see a perfect mate in 30,40 or more) but i believe that the time spent to them could be used more productively if it was used to improve the program's evaluation or search.... Contrary to that big opening books normaly have a huge influence on chess games. Actually I have an idea, where the difference in SMIRF's playing strength concerning 8x8 and 10x8 could be caused. Yes of course opening books is a whole different matter. Opening books help a program VERY VERY much to achieve good results..... Walter Mantego You're wrong if you don't think that's amazing to have such a big difference. It depends on how do you mean the amazing..... Take those data bases away from those machines and get the program down to SMIRF's size. Which machine is going to win? The endgame tablebases doesn't determine the strength difference at all! Period!!!!!!!!! The opening book is a huge factor that influences the strength diefference by much! The size of the EXEcutable of a (CRC) Chess program/engine doesn't matter! (Fritz 8 for example is 480 KB while Fritz 9 is 436 KB but Fritz 9 is way superior). We have that Gothic Vortex is 7 MB (many of these MB are just unused) and that its opening book is another 7 MB and its tablebases are 10 GB. So if you believe that all these 10GB - 14MB of tablebases do make a difference then you ARE ABSOLUTELY WRONG! Also why someone has to go down to Smirf size to compare against it? If someone has done all this work and if he has created for example a huuuuuuge opening book then why do we have to erase all these and test it without all these....? If you speak about _pure_ engine strength without endgame tablebases(althouth these don't add to strength at all at 10x8 variants) and without opening books, then yes we have to test the engines without endgame tablebases and without opening books! But if you speak about engine strength generally that means to compare 2 engines, then no we have to test the engines with endgame tablebases and with opening books and with whatever else their authors have thought about increasing their strength! How do you know that's not the reason? I was just speculating as to why it doesn't seem to play blitz games well, but you say it as if you're an authority on it. I know because i've observed its evaluations countless times and i have understood that it's one of these "slow searchers"..... This other program is not SMIRF, so why should it's weakness in blitz games have anything to do with SMIRF's? It the author's note telling you that SMIRF works in the identical manner? I didn't say that it has anything to do with Smirf! I just gave it as an example that there are other engines that can't play good at blitz while they CAN play good at longer time controls.... 31. Mars 2006, 22:28:55 Sujet: Re: Pythagoras: You're confusing me. You say the bases don't matter and thenyou say they're very important to a program's success a couple sentences later? Which is it?
"Yes of course opening books is a whole different matter. Opening books help a program VERY VERY much to achieve good results..... " and "We have that Gothic Vortex is 7 MB (many of these MB are just unused) and that its opening book is another 7 MB and its tablebases are 10 GB. So if you believe that all these 10GB - 14MB of tablebases do make a difference then you ARE ABSOLUTELY WRONG!" To me end table bases are the same as opening ones. They're just tables to look up the moves. They're not an example of the computer thinking. Just because a very large portion of these books aren't used during doesn't mean they are not a part of the program's strength. Take these opening books away from Vortex or the others that use them, will they beat SMIRF? 31. Mars 2006, 22:50:48 Sujet: Re: modifié par Chicago Bulls (31. Mars 2006, 22:53:04) Walter Montego: .
. . Yes you are confused.... -Endgame tablebases with 3,4,5,6 for example pieces(these are the endgame tablebases that have been generated until now, although i believe that 6 piece tablebases haven't been completed yet(for Gothic Chess)) are tables that programs look in their search and get perfect information about the position. For example if they end one moment in their search to have a position with 5 piece and they have/support endgame tablebases, then instead of wasting time to evaluate the position(that means with not sure results since their evaluation maybe wrong sometimes) they will have the perfect information (win,draw,loss) immediately! BUT in order these positions to occur in Gothic Chess or CRC we have to go to many captures so only few pieces to remain. But mooooost times the game is decided much earlier..... So if a program uses them at Gothic Chess i believe that only in 1 out of 400+ games there could be an advantage.... So we are speaking about zero advantage by using 3,4,5 or even 6 piece endgame tablebases.... -Opening books are tables that programs look at the beginning of the game and they represend knowledge gathered by millions of games around history. So they are priceless! So when i speak about bases i speak about "opening books" and "endgame tablebases". The first help a lot the latter doesn't. Take these opening books away from Vortex or the others that use them, will they beat SMIRF? It doesn't matter! The fact that Smirf doesn't use them, why should make other programs to adjust to what Smirf uses and what not? Every program uses what its author managed to give it as supplies to make it play better....! 1. Avril 2006, 00:18:07 Sujet: Re: Pythagoras: You have completely missed my point. The elegance of the solution. If we say that table look up and the use of books isn't allowed, then I say SMIRF is the superior program. You are being a results merchant. If every move in a game was known, the table look up method would work to perfection. Yes, yes, you've already proved there's less atoms in the universe than possible moves so there'd be little chance of having a way to store or access such a table. But supposing it was possible to have such a book of every move. Now compare this to a bunch of rules that tell when where to move that has maybe a 100,000 lines of instruction. Which would you say is the better solution? This is what I'm getting at. To me, anybody can use these table look up methods, opening, middle, or end game. To have a program not do this and still play well is to me an amazing thing and it is a program that I'd rather play against. What's the point of playing some machine that just looks up its moves? I might as well go get a book and look up for my moves too. It just becomes a matter of who has the better or larger data base. What's the point except that you'll win every time? When that happens it is no longer a game that is being played.
Programs that use an opening book is indeed a very helpful thing. I think that if all bits used or available to be used are counted, the program that uses the least is the best. Reinhard says 60,000 bytes. Both of you say Vortex has 7,000,000 for its program, another 7,000,000 positions for its opening book, and 10,000,000,000 for its endgame table. By my way of looking at it, SMIRF is the better program if it could play as well as Vortex. So maybe it can't play blitz Chess, but how about Embassy Chess at tournament time controls. What's that you say? Vortex can't play Embassy Chess? Or it's opening book is useless in Embassy Chess? Then SMIRF is by far the better program just by being versatile. I know which of the two I'd rather have. Can Vortex play the CRC as well as SMIRF? This is what I mean, SMIRF is able to play these games equally well without the use of books. Sure, if the goal of the problem is to devise the best playing program by whatever means at one's disposal, then yes, the program that wins the most is the better program. It was my understanding of SMIRF that Reinhard purposely did not work with that goal in mind, but instead wanted to create a program that played as it does. 1. Avril 2006, 03:14:40 Sujet: Re: Pythagoras: After reading my own post I'm thinking that perhaps I have it wrong. Winning the games is still important. Still, the goal of the programmer is also something to judge his result by. |
Date et heure
Amis en ligne
Forums favoris
Associations
Astuce du jour
|
|
Copyright © 2002 - 2024 Filip Rachunek, tous droits réservés Retour en haut |