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

Only Chess Forum

  1. 02 Sep '09 18:21
    Just thought I'd share an article that I saw on ChessBase News - About whether chess programs can be written to appreciate beauty in chess combinations. (I'll bet Greenpawn will have an opinion on that one! ) There's an attached PHD thesis that will take me a little while to wade through (over 300 pages).

    http://www.chessbase.com/newsdetail.asp?newsid=5730
  2. 02 Sep '09 18:54
    Originally posted by Mad Rook
    Just thought I'd share an article that I saw on ChessBase News - About whether chess programs can be written to appreciate beauty in chess combinations. (I'll bet Greenpawn will have an opinion on that one! ) There's an attached PHD thesis that will take me a little while to wade through (over 300 pages).

    http://www.chessbase.com/newsdetail.asp?newsid=5730
    very interesting article, thanks for sharing.

    fritz's calculation exercise has a system that gives points for your moves. the harder the move, the more points you get, and I think it's pretty successful on rating the difficulties of your moves. maybe that's a start
  3. 02 Sep '09 20:30
    Hi M.R

    I've down loaded the PDF and will look at it.

    It's sounds trying to get a computer to paint a picture ot compose
    a piece of music.

    Hi PP

    I cannot see how a computer can determine what is 'hard' for a human
    to solve.
    Is it based on depth of move or material sacrificed?

    It one of the things a computer cannot do (as yet) is to play a move
    knowing it's not the best but the refutation is hard for a human to spot.

    I've always been interested in this. Can you post an example.

    Cheers.
  4. 02 Sep '09 20:43 / 2 edits
    Originally posted by greenpawn34
    Hi M.R

    I've down loaded the PDF and will look at it.

    It's sounds trying to get a computer to paint a picture ot compose
    a piece of music.

    Hi PP

    I cannot see how a computer can determine what is 'hard' for a human
    to solve.
    Is it based on depth of move or material sacrificed?

    It one of the things a computer cannot do (as yet) is to pla human to spot.

    I've always been interested in this. Can you post an example.

    Cheers.
    I guess it rates the difficulty by the time it takes the engine to see the move, but I don't know. maybe it's a combination of depth&time. engines usually do spot difficult moves in a longer time.

    yes, engines cannot play such moves you're talking about, but they can be programmed to have tal-like styles if that counts.

    I had created a chessmaster personality that plays a totally sacrificial style. (I had called it Tal-Wannabe). It sacced pieces all over the place for activity and mobility, and in almost every game your kingside got ruined. I think it played around 2100 level.
  5. 02 Sep '09 21:20
    Hi

    I seen the machines where you can program in styles. had a bit a fun
    with a few of them myself.

    What they need is a program that has the ability to 'switch' styles mid-game
    to suit the position. A bit like a human has to do.

    On another thread the subject of K+Q v K+R came up and one lad
    said it was easy to beat a computer with the Queen.

    I believe him as the box can be so good it can see the Rook going
    and then all the way to checkmate so it may be playing moves not
    to save the Rook but to delay the mate it can see.

    I was thinking of using the 'settings' option to make the Rook the
    most important thing on the board - if possible put it on par with
    checkmate.

    The challenge the for the human is to win the Rook.
    I suspect this may it be tougher.

    Right I'm off to read that article,
    It's quite long so I'll be away for a day or two.
  6. 03 Sep '09 07:12
    Originally posted by greenpawn34
    Hi

    I seen the machines where you can program in styles. had a bit a fun
    with a few of them myself.

    What they need is a program that has the ability to 'switch' styles mid-game
    to suit the position. A bit like a human has to do.

    On another thread the subject of K+Q v K+R came up and one lad
    said it was easy to beat a computer with the Queen.
    ...[text shortened]...
    Right I'm off to read that article,
    It's quite long so I'll be away for a day or two.
    [b]On another thread the subject of K+Q v K+R came up and one lad
    said it was easy to beat a computer with the Queen.[b]

    I think this is because the computer always finds the move that is furthest from checkmate creating a pattern you can learn to beat it with. A similar thing happens with the Knight & Bishop checkmate.

    Deviating from the pattern (as a human may do) might provide a shorter path to checkmate but takes the attacker off piste.

    Learning these endings reminds me of the rubics cube.
  7. 03 Sep '09 10:14
    Started on the PDF file - there is a lot of woffle.

    I jumped ahead and found the first chess diagram does not appear till page 102.

    I'll stick with it but I am looking for excuses NOT to read it.

    Do these thesis things have to be over a certain length. the guy appears
    to be using a page to say something when a simple sentence would do.
  8. 04 Sep '09 11:14 / 2 edits
    Originally posted by greenpawn34
    On another thread the subject of K+Q v K+R came up and one lad
    said it was easy to beat a computer with the Queen.

    Yes, that's a problem on Chesstempo too, which causes a disadvantage for high rated players.

    a 2000+ rated player sees the position and very quickly recognizes the opportunities and candidate moves, and starts to calculate. They usually try to calculate (I always do) every line to it's conclusion and it's a long and tiring job, as there are usually very tricky lines you must see to be exactly sure if the move works.

    The casual 1600 rated player also recognizes the candidate move and plays it very quickly, without going deep enough to come out of complications clean in his calculations.

    you make the move, and the computer responds ...queen takes bishop (a move a human would never make, but the computer can't leave it's objectivity of course) and problem is over, both have answered the puzzle "correctly".

    although this is something to rant about, it's not too much of a deal, it happens rarely, and it can't be fixed anyway.