Chess - Play & Learn


FREE - In Google Play

FREE - in Win Phone Store


Does making a chess engine make you good at chess?

  • #41
    NimzoRoy wrote:

    Is someone a better driver because they build cars for GM or Ford? 

    Bad comparison. Could someone become a better driver by programming a driving AI? Maybe so. When you program ANYTHING, what you're learning how to do is codify your knowledge. How to take a human's knowledge and put it into a super solid form that a computer can understand, without any "feel" aspect. You don't program "stay within the lines," you program a function to measure the distance from the lines and then either split the difference, hug a set distance from the left or right, etc. Plus this has to be overridden whenever you change lanes. The program is concrete, but still has to behave rationally, as most humans do, and there is the challenge. 

    Honinbo and irobeth: good points. Most use alpha-beta search, which is why no engine can find the "correct" move on sprung halosar trap (at least not when I tried 2 years ago). It writes off (prunes) the queen sacrifice. Writing a better evaluation function would indeed make for a better gameplay, assuming the calculations could be done far enough down the tree. Playing 20 moves out is useless if you miss the "good move," but sometimes you can't tell the good move till 5+ out. 

    Humans don't brute force: Chess engines do. Chess engines play for straight material or board position points (largely) while humans can sacrifice pieces better. I think if chess engines can be made more effective through a better eval function they'd be able to actually calculate less (pruning more- though depth 1 pruning is why they miss halosar!) so it's a balancing act until we get large enough computers to solve chess.


Online Now