[Melbourne-pm] Timer::HiRes and alarms? (Was: AI Contest)
Kim Hawtin
kim at hawtin.net.au
Thu Nov 11 15:32:05 PST 2010
Sam Watkins wrote:
> You definitely would need to use an algorithm that makes the most of its time,
> because the more it thinks the more chance to find the best move, unless the
> game is trivial.
Part of my 'strategy' has been to spread the computation across several
turns. does it work ... sort of.
> I'm not sure how the top bot could be a Lisp bot - perhaps that Lisp hacker is
> very very clever! or perhaps they are using a very neato fast Lisp compiler.
It is; SCBL. An old one, but none the less, probably the fastest free
compiler.
> I suppose to win, you'd need to do game analysis like a chess-playing program:
> - tree search, mini-max pruning, etc, and throw in some good heuristics!
Certainly treat it like an eigan vector/value pair and fiddle with the
coefficients and use a number of working sets. Thats where I'm headed.
The main issues that my strategy fails down on at the moment, is a more
openly spaced map or where there is few larger growth planets near the
home planet. So part of the first turn is to pick a set of coefficients
that work in that context...
And no, I haven't updated my bot on the site in some time =/
cheers,
Kim
More information about the Melbourne-pm
mailing list