[jukofyork]

[ QUOTE ]
[ QUOTE ]

(1) Surely there must be a better way to do this than simply assign each player a hand independently from their distribution, then check if any of them have been assigned the same cards and keep restarting until a valid configuration is reached?

[/ QUOTE ]

here's an idea that may be feasible depending on the application.

use method 1 as above, but if you're finding a large number of "collisions", forget monte carlo and enumerate all the possible combinations and apply weights accordingly.

this should work since in practice, if you're getting a lot of collisions, you're probably selecting from a small number of hand combinations anyway and enumerating them all shouldn't be too much work.

in some extreme cases, this wouldn't be particularly feasible - but it would probably work well in practice. (examples when it wouldn't work: if there's 10 players with all having wide ranges, or if there's 3 players who all have AK 99% of the time and any of the 168 other hands 1% of the time)

[/ QUOTE ]
Hi, sorry for the slow reply - been away for a few days, then had problems logging onto 2+2 until I cleared my IE cache and cookies.

Yep, I agree that enumerating is prolly a good idea for the cases where the ranges are very constrained and possibly even adding something too do it after so many failures (I think detecting an impossible configuration before you run though is going to be O(n!), as you will has to test all possible player orderings to be sure).

I haven't had time to do much more with the code since my last post, but overall it seems to be working quite well and sometime over the next few days I'll write a simple CLI app to let you input a hand vs a distribution of opponent hands and maybe somebody will find it useful and/or extend it to do something more useful.

Juk [img]/images/graemlins/smile.gif[/img]