Author: Christophe Theron
Date: 10:03:18 10/11/00
Go up one level in this thread
On October 11, 2000 at 03:03:19, Uri Blass wrote:
>On October 11, 2000 at 02:09:05, Christophe Theron wrote:
>
>>On October 11, 2000 at 00:32:23, Uri Blass wrote:
>>
>>>On October 10, 2000 at 21:27:38, Christophe Theron wrote:
>>>
>>>>On October 10, 2000 at 20:47:28, Vincent Diepeveen wrote:
>>>>
>>>>>On October 10, 2000 at 16:16:34, Bruce Moreland wrote:
>>>>>
>>>>>>On October 10, 2000 at 10:52:54, Pekka Karjalainen wrote:
>>>>>>
>>>>>>>
>>>>>>> Hi folks,
>>>>>>>
>>>>>>> there were some postings some time ago by Bruce Moreland, I believe, about
>>>>>>>tablebases for chess on the 6x6 board. That seemed quite interesting actually
>>>>>>>and I started to wonder: Could we solve 6x6 chess? If not, how about an even
>>>>>>>smaller variant?
>>>>>
>>>>>it's nice to talk about small chess at 6x6, but some pieces you
>>>>>gotta miss then which you have on 8x8. like if you remove
>>>>>the queen and the king, then the game gets totally different.
>>>>>
>>>>>how do you want to castle without rooks for example, not to mention
>>>>>without king?
>>>>>
>>>>>if you remove the queen and another piece on each side then you have a
>>>>>pretty easy positoin left to play. how the hell do you plan to
>>>>>win that anyway?
>>>>>
>>>>>within a few moves all pawns are against each other and you shake hands
>>>>>and it's a draw.
>>>>>
>>>>>somehow i'm missing realism here.
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>>I don't believe that we can solve 6 x 6 chess via retrograde analysis, since
>>>>>>powers of 36 are not shockingly better than powers of 64.
>>>>>>
>>>>>>I built a 6-man table and it was well under a gigabyte, but if you try to add a
>>>>>>seventh you are doing multiple gigabytes again, and there are 24 pieces on a
>>>>>>full board.
>>>>>>
>>>>>>In case anyone has a burning desire to know, there aren't any huge long wins in
>>>>>>KRR vs KRN on a 6 x 6 board. This s contrasted with 8 x 8 chess, where there is
>>>>>>some godawful huge conversion case. It seems intuitively obvious that a knight
>>>>>>is stronger on a 6 x 6 board than it is on an 8 x 8 board, and perhaps this is
>>>>>>enough to draw if there isn't a way for either side to win immediately.
>>>>>>
>>>>>>You must be right that it would be possible to solve chess on a board that's
>>>>>>small enough. I haven't tried to figure out what that size would be. It might
>>>>>>be practically impossible to do it for any interesting case.
>>>>>>
>>>>>>bruce
>>>>>>
>>>>>>>
>>>>>>> You can find several actual small chessvariants from this URL:
>>>>>>>http://www.chessvariants.com . Basically you can make them up yourself by
>>>>>>>taking a smaller board (any size from 3x1 upwards is possible). Just set the
>>>>>>>starting position as you please and call it <foo>chess. The smallest ones are
>>>>>>>obviously trivial and can be solved even on the back of an envelope. You might
>>>>>>>want to remove the castling rule and double pawn moves, and even treat stalemate
>>>>>>>as a win to avoid total drawishness.
>>>>>>>
>>>>>>> But this is not what I am after really. I was thinking that would it be
>>>>>>>interesting to try to solve these? Could we get a program that would search all
>>>>>>>the way from a starting position to its (smaller NB) tablebases?
>>>>>>>
>>>>>>> One might even get a little feeling about the actual (and huge) computing
>>>>>>>resources that would be needed to solve standard chess. If 5x5 takes this much
>>>>>>>effort and 6x6 takes that much then 8x8 takes SO much (exponentially more of
>>>>>>>course). What would a solution to a 5x5 chess variant really look like?
>>>>>>>
>>>>>>> As I am not much of a chess programmer I do not want to take on this challenge
>>>>>>>myself. But if it interests anyone at all, maybe it was worth mentioning.
>>>>>>>
>>>>>>> Comments?
>>>>>>>
>>>>>>> Pekka K.
>>>>
>>>>
>>>>
>>>>The first chess program that actually ran on a computer was on a 6x6 board (the
>>>>bishops had been removed).
>>>>
>>>>I think it was called the "Palo Alto Chess Program".
>>>>
>>>>Some time ago I have downloaded a program that plays this game. The game is less
>>>>interesting than chess of course, but it is still entertaining, and you've got
>>>>some new problems to deal with.
>>>>
>>>>For example the king becomes a very embarassing piece in this game (because
>>>>castling is not allowed). Wherever you put your king, it always seems to be in
>>>>the way of some other useful piece. Funny.
>>>>
>>>>BTW after a few attempts I have been able to defeat badly the program in
>>>>question. 6x6 chess must be much simpler for the human brain.
>>>
>>>I think that the reason that you could beat the program is the fact that the
>>>programmer did not work on the program like programmers worked on chess
>>>programs.
>>
>>
>>
>>I wonder. The program was easily reaching 5 to 6 plies and I beat it.
>>
>>If I play against a chess program looking 5 or 6 plies ahead, I always lose.
>
>This is not the right comparison because it is also more easy for computers to
>search 5-6 plies in 6*6 board.
>
>I also think that the evaluation function and the extensions are important.
>
>>
>>I guess my victories are related to the fact that on a 6x6 board the number of
>>plans is rather limited. Also the human understands quickly how to block the
>>position so the opponent pieces cannot play, then how to attack the opponent's
>>king.
>>
>>
>>
>>
>>>I believe that 6*6 chess is easier than 8*8 chess also for programs.
>>>I think that increasing the board is going to give more problems for computers.
>>>I know that the game go is hard for computers because the board is 19*19.
>>>
>>>I believe that doing chess programs for this size of board will force
>>>programmers to use selective search if they want to have good programs so maybe
>>>you and Ed have an advantage if you decide to do it because I understood that
>>>you use selective search.
>>
>>
>>
>>I would not pretend that we are the only ones to have very selective searches.
>>
>>I think that The King, Hiarcs and Genius are also very selective (but not
>>similar).
>
>I think that being selective in the root is important in 19*19 board.
>I read that you are selective in the root.
>I believe that hiarcs,Genius and chessmaster are not selective at the root but I
>may be wrong.
>
>It is possible to check it.
>Programs show the move that they consider.
>
>It is possible to watch programs for some hours,write the moves that they
>consider and see if they consider all the moves or only part of them.
>
>Assuming that programs are not lying about the move that they consider you can
>find if programs use selective search in the root.
>
>Uri
No you can't. Some programs, like Chess Tiger, refresh the screen only once per
second. So if a move is considered for a short enough period of time, you won't
see it. But it has been considered.
Also, I did not say I was selective at the root. I said the 16 bits version of
Chess Tiger was. I have said nothing about the current version.
Christophe
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.