Please turn on javascript in your browser to play chess.
Only Chess Forum

Only Chess Forum

  1. 18 Dec '11 10:16 / 1 edit
    It's been discussed before, the theoretical possibility that some ultra future supercomputer could play perfect chess.

    Some computer programs will analyze your games and tell you how many mating opportunities you had (if I win, I often miss earlier mating opportunities that I didn't see.)

    It dawned on me that it's entirely possible that after the first two moves are played, either white or black could already be caught in a mating web. It's just that no human or computer has the capability to see it. Of course this web would be so incredibly convoluted that there is an astronomically large number of ways to get there, but if you know the moves there is literally nothing your opponent can do to stop mate.

    It's also theoretically possible that the game starts with black already caught in a web. Maybe after 1. E4 or 1. D4 black's demise is inevitable. It would be hilarious if the killer blow was something crazy like 1. H3!!
  2. 18 Dec '11 11:59
    Maybe white begins in zugzwang
  3. Standard member TimmyBx
    TacticsTime.com
    18 Dec '11 12:26
    well I think that this type of thing coming about isn't going to be because of a super computer, but lots of computers creating/solving and storing the various tablebases - where individual positions are solved out to mate, or a forced draw.

    For example, checkers has been solved. A very weak computer can play "perfect" checkers, just as long as it can access this data. It is not thinking or calculating, just looking up positions, where it can find the best move.

    http://en.wikipedia.org/wiki/Endgame_tablebase has some interesting information about it.

    I recently installed all of the 3,4, and 5 man tablebases, and it is very nice for doing analysis of endgames. I believe that all of the 6 men tablebases have been solved as well, and quite a few of the 7 man tablebases.

    Once we have a 32 man tablebase, chess will be solved :-)
  4. 18 Dec '11 12:54
    Originally posted by Robin Johnson
    Maybe white begins in zugzwang
    Haha! Wouldn't that be something.
  5. 18 Dec '11 14:03
    Originally posted by USArmyParatrooper
    Haha! Wouldn't that be something.
    Or both, and the game is drawn because of the repetiton line: 1Nf3 Nf6 2.Ng1 Ng8 etc....
  6. Standard member ChessPraxis
    Cowboy From Hell
    18 Dec '11 16:45
    Originally posted by Mephisto2
    Or both, and the game is drawn because of the repetiton line: 1Nf3 Nf6 2.Ng1 Ng8 etc....
    White to move, mate in 164. 🙂
    Helpmate in 148
  7. 19 Dec '11 03:17
    There is no perfect chess. Period. The game is too complex and there are people that are too creative to stop advancing.
  8. 19 Dec '11 10:02
    Just an idea:

    Endgame tablebases contain data about each possible position:
    - whether perfect play results in win/draw/loss
    - the way (alternatives) to get there (but this is not that important for the analysis I have in mind)

    Apparently, these tablebases exist for positions up to about 6 pieces, because the more pieces the more complex. Nevertheless, we now there is a deterministic answer for each position. Because each 6 piece position results from a 7 piece position, in which there were finite number of choices. So there is the theoretical possibility to solve the initial position. (see remark)

    Now, I'm wondering what we could learn from looking at the relative fraction of W/D/L for each type of position (nr of pieces). Does the relative fraction of draws increase with the number of pieces? This may give some hints about what we can expect from the initial position. But there is still a long way to go before we get to that point. Can this data be found online?

    Remark: When we get to about 25 piece tablebases, I can imagine that not all positions can be legally reached (e.g. white pawn on e5, black pawn on e4). This may be important, because they skew results and cost analysis time. It's easy to check whether a position is illegal, but whether it can be reached legally is more difficult. Are there ways?
  9. Subscriber Marinkatomb
    wotagr8game
    19 Dec '11 13:20
    I'm not so sure it can be solved myself. I think 'perfect' play results in a draw, simple as that. For one side to lose it requires a mistake or two. There is also an issue with style, while a lot of positions lend themselves to pure calculation (which is computer territory) but there are always positions that are compkletely balanced and can be aproached in a number of different ways (or styles if you like). To be able to calculate every possible move to mate from a theoretically level middle game just seems impossible if you ask me. Sure a computer will beat a human in such situations, but two equal computers running the same software...impossible.
  10. 19 Dec '11 13:55
    I simply mean that the game is deterministic. The practical realization of solving chess is a long way ahead or may even never be realistically possible.

    By the way, also a draw could be the 'solution' of chess. It's probably about half of the people that think that white wins and half that it's a draw. Probably also a few that think that black wins.
  11. 19 Dec '11 14:35
    Originally posted by Quits
    There is no perfect chess. Period. The game is too complex and there are people that are too creative to stop advancing.
    That's not true. There is a finite number of possible positions you can reach throughout the course of any game, which means there must be a perfect game.

    Currently it is too complex for any human or computer to calculate, but that doesn't mean there's no such thing as a perfect game.
  12. 19 Dec '11 15:45
    Suppose that tablebases were available for all 32 men. We know that this process may take longer than the expected lifetime of the universe (but shorter than the time to ban some cheats on RHP&hellip😉. Ok, so I’ve installed my 32-men tablebases from floppy disk onto my PC. Does this mean my PC is now the ultimate player? Close but maybe not perfect as such.

    Assume that the perfect game is a draw if White’s plays 1.d4 or 1.c4 (1.e4 loses to the Caro-Kann… wait and see). Now consider an opponent that is perfectly prepared for 1.d4 but not 1.c4, and another opponent that is vice versa. Here, the “ultimate player” would have to have knowledge of the opponent in order to maximise results. i.e. the tablebases alone are not enough.

    Seriously though, this does have implications for the current state of art. Current engines could take into consideration what may be most problematic for a human when analysing an endgame with tablebases. This could be based on measures such as the need to find “only moves”; heuristics based on what is not intuitive for a human; setting of traps; etc. Imagine tablebases that don’t just have “perfect play” data but also probabilities based on “human play”.
  13. 19 Dec '11 17:55
    I guess when that day comes we all switch to playing Go. Or spending time with our friends and families.
  14. 20 Dec '11 00:37
    Originally posted by USArmyParatrooper
    That's not true. There is a finite number of possible positions you can reach throughout the course of any game, which means there must be a perfect game.

    Currently it is too complex for any human or computer to calculate, but that doesn't mean there's no such thing as a perfect game.
    Only time will tell, but I don't agree with your statement.

    Creativity is an incredible thing, and chess is a very complex game that is continual changing from game to game and era to era.

    Many players have gone mad searching for the "perfect game".
  15. 20 Dec '11 02:33
    Even if you could determine every possible position it would then be a matter of which positions are better... which is what we do now. Conclusion; there is no need to know every possible position.