The question that prompted this is this one: I voted to close it as being too broad, specifically because of this line from the FAQ

If you can imagine an entire book that answers your question, you’re asking too much.

but also because a question about how to improve at chess can't be answered well in this format. (A good answer would take up at least a page.) However, I think that some aspect of this question does fit the site well ... I'm just not sure how best to do it while keeping the resulting question relevant for the OP.

The most recent edit seems to me to be heading in the right direction, but still doesn't supply a narrow-enough scope to the question. I think there are a couple of options that might work, but I'd like to get some input on them, as well as other possibilities, so we can get something workable out of this.

  • One option is identifying a particular part of the game where the AI seems to take the lead: "How can I improve my (openings|mid-game|end-game)?" This would be more of a "teach me to fish" question, looking for ways in which to improve.
  • Another would be to be more specific, showing an opening, series of moves, or position that seems to recur. "I'm having trouble with situation X against the computer, what can I do?" This is closer to questions on chess openings that are already on the site.
  • Still another would be closer to the original intent of the question: if the OP can identify specific features of the AI, or even the application and the difficulty setting, then the question could be "I'm struggling against Program X on difficulty Y; how can I close the gap?" One problem with this approach is that it could still be too broad; unless there's something specific about that AI, it could still be a version of "How do I get better at chess?" (Or alternatively, too specific because of the reference to a specific difficulty level ... maybe that version would do better on Gaming.SE.)

As has been discussed before, the original question is too broad, but there is a way to save the question, isn't there?

| |

The issue here is that the question is just not very good. It's not really the fault of the OP, who by his own admission is inexperienced, but the question is naive and the answers are not likely to be helpful.

It really doesn't make a lot of sense to me to talk about how to beat a chess computer. To try and identify a particular 'style' associated with chess A.I. in general, or with a specific chess engine, and then to work against that style, may be possible. It's just not a very good idea. There are only two ways to go with the question:

1) Talk about classic A.I. weaknesses. You could say something about how a weak A.I. tends to be poor at sacrifices and combinations, or rook endings, or obscure openings not in its opening book, and waffle a bit about search depth. Whether or not that applies to a modern A.I. on your Mac is probably going to be pure speculation. Whether you can usefully exploit it is speculation again.

2) Talk about general chess improvement. Practically, this is the better answer, since your time would be much better spent improving at chess, than at improving against some peculiar A.I play style. However, clearly, such an answer has nothing to do with A.I. It's also a huge, far too broad topic.

Some A.I. programs allow you to 'dumb down' their playing ability, so that you can have a more balanced game. That seems like the obvious fix, if winning is actually the goal.

A more general point is that improving at chess requires being regularly beaten by good play, so that one can learn and raise one's game. In that sense, the feeling of frustration and striving is rather like learning a musical instrument. If a player doesn't find anything intrinsically enjoyable about that, then chess probably won't be a particularly fulfilling game.

| |
  • Thanks for recognising my chess inexperience: I think in most games, it would be a valid assumption that a computer player would have obvious glaring weaknesses that it would be easy to attack into. Maybe not so much in chess, apparently! If good play can be based on sheer computational power and not really ever on imagination or finesse, then it's not the game for me. And my question has been answered satisfactorily by the closing :) – thesunneversets Sep 13 '11 at 10:12

Would it have been better if I'd approached it from the other angle? "In what ways is a computer's chess game deficient compared to a human being's?" Asking about specific weaknesses in a machine's game.

As opposed to my more nebulous "how do I beat a computer?", which of course has a million potential answers with the startpoint "first, get better at chess, like this:..."

| |
  • I think that's a more specific question and could lead to much more specific answers ... but the issue could be that unless you ask about a specific AI, as ire points out, we might not be able to give valid answers. What might work best is to split your thought into two pieces: one SE-acceptable piece that can be reopened, and one SE-unacceptable piece to ask in chat. Hopefully here we can get you the former. – Dave DuPlantis Sep 12 '11 at 18:19

As I understand the question, it was "How can I get to a point where I can beat/draw against a computer, specifically MY computer?"

A simple answer might be, buy a computer you can beat. If Gary Kasparov can't beat Big Blue, you can't either. But there are some computers that YOU can beat. Which ones are they?

One way of asking the question is, my strength is X. What are the specifications of a computer (in terms of memory, calculating ability, other attributes) with which I can have reasonably balanced games, and win or draw at least some of the time?

A second way to ask the question is, I have a computer with specifications A, B, c. How strong would I likely need to be in order to beat it? Does the computer have a peculiarity that would give me a chance even if my official strength is decidedly below its?

In a sense, it's not too different from finding a human opponent. My strength is something like 1500, give or take. In order to find "good" opponents, I'd be looking for someone whose strength is in the 1400-1600 range. I remember once playing "evenly" with a 1700 strength opponent who had a particular weakness I could exploit. (But this man held his own against other 1700 opponents who easily beat me.)

"How can I beat a computer?" is an overly broad question that Gary Kasparov can't answer. An acceptably narrow question might be about how would which man fare against which computer? Or in human terms, how might Tom Au fare against Dave DuPlantis or the thesunneversets?

| |
  • I don't think specs are really relevant any more: certainly for people new to chess, pretty much any program these days is capable of beating them handily. (A question along those lines might not be on topic anyway; I think we want to focus on the chess aspect of the problem as much as possible.) I think I like a variation of the last part of your second question best - how can I beat this level of the AI? - but I think we'd still need more detail about the AI itself to help, or it goes back to "How do I win at chess?" – Dave DuPlantis Sep 12 '11 at 18:24
  • @Dave: One variation of the question is, are there AI's that exhibit enough variation from game to game that they can be beaten by a theoretically weaker player (like my 1700 opponent)? And "I'm 1500 (no rank beginner), how I can find a computer program that's not tow challenging? might be OK. (I've found some, but they are mainly older ones.) – Tom Au Sep 12 '11 at 18:51

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .