🚀 go-pugleaf

In article <1992Oct7.133312.17295@odin.diku.dk>, rael@diku.dk (Peter Hj{resen) writes:
|> Hello world...
|> 
|> 
|> [This is my first posting, so bear with me...]
|> 
|> I have been following this group for quite a while, and I (almost) 
|> haven't seen one single discussion about RISK. I think this is a shame...
|> 
[stuff about what to do for distribution of windows risk game deleted]

post your source code (if you want) to comp.sources

if you're going to let people know about the program yourself,
any ftp site is as good as another. if want people to find it
themselves, submit to a common archive.

note, if you post it to comp.sources, it will get archived there
automatically

|> 
|> Right now, the program uses an ad hoc playing strategy, based on local
|> observations. I would like to hear from any of you, who migth have 
|> interesting views on the possibility of using 
|>    
|>      1) Alpha/Beta-pruning (minimax?).

if you're going to use minimax, of course use alpha/beta pruning
to reduce the search space, search deeper in the same time, whatever
tradeoff you choose. the big problem is the heuristic that evaluates
board position. what do you want to maximize (minimize)?

for instance. i've seen risk programs that clearly choose the country
against which they have the best odds each time. there is no strategy
for expansion.

your heuristic would clearly need to evaluate:
- expected number of points for armies next round
- probability of cashing in cards for armies on next round
- order of play (if you play first you need to be defensible,
  if you play last, you can take a stab at a continent just to
  take away the other's bonus)
etc.

this is always a good approach as long as:
- you can make a reasonable evaluation of board position
- you can do it fast enough to search far enough ahead
- the evaluation heuristic is not too obvious

keep in mind that you can't search very far ahead in a game
like this, and there's really no point. each player can make
large numbers of moves or none. the search space is too large.

having a table of combat loss probabilities for combinations of
attacking/defending armies makes combat prediction a simple lookup,
but how many armies will be moved in, from where, will he attack
again from there, etc?


--------------
Grey

my opinions, not those of Bell Northern Research