home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!charon.amdahl.com!pacbell.com!ames!agate!doc.ic.ac.uk!uknet!rsre!signal.dra.hmg.gb!signal.dra.hmg.gb!usenet
- From: heading@ccint1.rsre.mod.uk (Anthony J.R. Heading)
- Newsgroups: sci.math
- Subject: Re: request for info on game theory
- Message-ID: <1k6imqINNer5@signal.dra.hmg.gb>
- Date: 27 Jan 93 18:02:02 GMT
- References: <1993Jan23.212705.287@ctsx.celtech.com>
- Organization: Defence Research Agency
- Lines: 46
- NNTP-Posting-Host: amigos.dra.hmg.gb
-
- davidd writes
- >
- > Hello, I'm a new reader to this news group and was wondering if
- > anyone could offer any advice. For some years now I've had
- > various ideas for a computer/board game. In a nutshell, (in the
- > computer version) you would play against a autonomous agent. (no,
- > not a shoot them up thing) Anyway I was talking to a friend over
- > lunch the other day and was babbling incoherently about some of
- > my ideas and she was wondering what game theory books I had
- > studied or used. I hadn't. I didn't really know anything about game
- theory
- > as such so I thought I'd better look into it. If anyone could
- > offer any recommondations, I'd appreciate it. Hmmm, just a
- > thought, perhaps should I be looking for any relevent AI
- > book...??? Thanks for your time and help.
-
- I'd recommend reading Hofstader's (?spelling) book
- "Metamagical Themas" - it gives rather a good overview of his
- thought processes in going through trying to do what you want
- to do, and explains concisely the strengths and weaknesses of
- game theory. As well as lots of other funky stuff.
-
- On a more conventional academic level, the definitive book
- is Von Neumann and Morganstern's "Theory of Games". In it they
- invent game theory. So it'll tell you all you need to know, but
- you have the tedium of working out the weaknesses for yourself.
-
- Effectively, in a sentence, game theory is about going 'all out'
- for a draw. You assume your opponent is going to play optimally,
- and play to block him as best you can. Your move is to minimise
- the opponents maximum gain. In a symmetric game, this tends to
- lead to a draws - however clever you are, a game-theory based
- opponent will block you: not try to win, simply try not to lose.
- It'll have considered every move you could make, and play to
- make sure that none of them are a real winner for you.
-
- OK, this is fairly good - but it wouldn't work very well if your
- opponent is a bit predictable. If you can model your opponent -
- read his hand as it were - and bluff a bit, you stand a chance
- of beating him. This is a consequence of the way people play games.
- A game-theoretic opponent is unreadable - completely random within
- its prescribed strategy. In fact, while it's normally an advantage
- to know your opponents strategy - with a game-theory based opponent
- the strategy can be laid open and bare, and it just doesn't help.
-
- Anthony
-