13

I've been playing a lot against an old build of Chess Free on Android the other day and noted that the AI almost always chooses the safest course of play, boringly eeking out positional advantages. If you play at its pace and react to it your games are dull. However I've got a feeling that if I play this type of game but avoid exchanges I can fashion opportunities via a series of sacrifices and exchanges that it seemingly hasn't considered. I'm guessing the computational cut-off (60 seconds in this case) prevents the AI from considering these sacrifices once the complexity of all pieces being developed arises.

Are there any other ways of abusing the fact that most chess AI's just brute force and typically go with the "safest" line?

4

Five ideas in my head to win against AI's chess:

  • trying to play unusual openings
  • better knowing theory than AI for several openings
  • trying to play avant-gardist positions (sacrifices and planning move in advance)
  • analyzing games of AI to determine the way it plays
  • trying to know how AI has been programmed

You have tips here and interesting information there.

  • 2
    I don't think that well-known mates will be possible, since the computer will "see" these mates. I also would avoid avant-gardist postions, since they can be full of (tactical) surprises and you cannot surprise a computer. Computer normally have huge opening databases. It does seem easy to know more than a computer – Michael Mar 1 '13 at 16:38
  • Those links are really interesting, especially the hippopotamus which strangely enough was my favourite opening when I was about 7 years old :D. – Quibblesome Mar 1 '13 at 16:39
  • @Michael: I have edited my answer. – Zistoloen Mar 6 '13 at 20:43
3

The computer is very good with tactics and bad with strategy. There is little change to outsmart a computer in tactics. So try

  • closed strategic positions
  • exchange queens early
  • seek the end game
  • Can you elaborate on the difference between tactics and strategy? Personally I would expect the computer to have the advantage after exchanging queens and the end game because it plays into the fact that the machine can compute the remaining possibilities. I would posit that keeping pieces on the board extends the possibilities thereby suggesting that the machine doesn't have enough time to consider all of them. AIs are still essentially brute force aren't they? – Quibblesome Mar 1 '13 at 16:37
  • end games are often technical. You don't need to calculate much, but you need to know what to do. Humans who know what to do can "see" many moves in advance and avoid the calculation of useless variants. You can much easier know what position is a draw or a win without thinking about it. You cannot compute better than the computer. Making the position easier (and keeping it closed) helps you not the computer. Quiet positions and long-term planning can make you foresee situations with little effort that the computer cannot calculate – Michael Mar 1 '13 at 16:43
  • @Quibblesome: you can count and see that your pawn needs 4 moves to promote while his needs 5, but the AI needs to see millions and millions of positions to conclude that. May well be beyond its horizon. There are more pieces in the middlegame, but fewer moves that they can make because the positions are less open. On the other hand, the chance that you miss some hidden tactic is really large in the middlegame. – RemcoGerlich Mar 3 '13 at 10:06
  • 1
    @Quibblesome simply put: tactics is short-term, strategy is long-term. Your reasoning about the endgame is fair, but it ends up not working quite like that. For whatever reason, humans have very good heuristics for endgame positions. E.g, we can distill a position into "as long as I can keep opposition, it's a draw. AIs typically have a limited set of endgame heuristics like this (although it's improving all the time). (continued below) – Daniel B Mar 5 '13 at 7:45
  • This, combined with how a game-tree branches, means that making the position more predictable along broad lines favours humans (basically, our heuristics for straightforward endgames is still better than the AI's). Taking the queen off the board tends to make things a lot more predictable. – Daniel B Mar 5 '13 at 7:47
3
  1. Playing an unusual opening can reduce the depth of the searches made by the computer in the middle and end game in a timed match.

  2. If you avoid exchanging off pieces unnecessarily then the computer has to do larger searches which will be to a shorter depth.

  3. Think about sacrifices that might benefit you if the computer accepts them and doesn't search that branch far enough to see what your plan really is. Be careful with this though as some computers will continue to search all branches.

  4. Against better computers, the computer will be using your move time to think about it's reply so the longer you take to decide on your move then the longer the computer has had to calculate the best reply.

  5. Unless your end game is excellent, avoid end games whenever possible as Chess computers can do extremely large searches in the end game.

  6. Avoid repetition of moves as this allows computers which use transition tables to save a lot of time as they can move instantly when they already have the position and best moves to play stored in a transition table.

Found here: http://www.becomeawordgameexpert.com/computers.htm

  • 1
    Same link I have given in my answer. Please read other answers before giving your opinion. However, you give some good points. – Zistoloen Mar 2 '13 at 16:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.