🚀 go-pugleaf

In article <1992Aug13.133407.12508@exu.ericsson.se>, exuhag@exu.ericsson.se (James Hague) writes:
|> 
|> I've found that for "most" games (excluding pure strategy games), you
|> can get by with a "fake" computer opponent.  That is, the computer doesn't
|> use sophisticated AI routines, but rather uses only a few simple rules with
|> a little randomness thrown in.  You can make the computer act more
|> aggressively by tweaking a few values.
|> 
|> The more complex a method used, unless you go full out with an exhaustive
|> search tree (not possible for many games), the higher the chances of
|> there being holes in the computer's strategy, IMO.  Unless you plan to devote
|> most of your time devloping AI algorithms it is better to go with something
|> simple and use speed, number (i.e. multiple attackers at once), and randomness
|> to overload the player.
|> 

My understanding was that genetic algorithms would be done once, and they 
would tweak the values for you so that you don't have to guess.  You could
run the GA every night, and in a few weeks, you'd have such a good opponent
that it could beat any other possible opponent using those few simple rules.
(It wouldn't be done at run-time, so it wouldn't slow down the game...)

	Amit


-- 
/\\ Amit J Patel, amitp@owlnet.rice.edu
\\/ I think I'm at Rice University