A comforting thought
“My insights into Fairy Tale strategy are markedly better than random play.” OK, it’s not that comforting.
Based on feedback, I’m skipping the Neural Network and just going with a correlation matrix (which will eventually evolved). I’ve hardcoded my default correlation matrix (80 x 40) and initial weighting of cards (1 x 40). [Player #1 is the non-random one].
INFO 9859 - GameState - Player #1 won 82 games. INFO 9859 - GameState - Player #2 won 9 games. INFO 9859 - GameState - Player #3 won 5 games. INFO 9859 - GameState - Player #4 won 8 games. INFO 9859 - GameState - Player #5 won 3 games.
Right now I’m just using a “Play the card you like best each round”, but perhaps I should look at the breakpoints … if I know I won’t be playing two of the cards at all, perhaps theirs a way to let it evolve ways of ordering the rest of the card play.
Now I just have to program a mutator, a serializor and a reaper. Then it’s off to the evolutionary races! (Eventually I’ll need to write a GUI so I can play against it, assuming it gets good enough …)
I’ll write some more bridge in a few days. And some actual game stuff later…
Brian – ran across this – it might be useful in your development:
http://xkcd.com/534/
SeanP
May 6, 2009 at 10:02 am
Very cool. I assume you are playing fairy tale with the advanced cards?
I think that non-partnership fairy tale is a pretty simple score maximization excercize, once you know the cards. Partnership is much more interesting because you must weigh defensive vs offensive moves.
Alexfrog
May 6, 2009 at 12:50 pm
Very cool. I assume you are playing fairy tale with the advanced cards?
I think that non-partnership fairy tale is a pretty simple score maximization excercize, once you know the cards. Partnership is much more interesting because you must weigh defensive vs offensive moves.
Alexfrog
May 6, 2009 at 12:50 pm
Also, as SeanP noted, make sure to add that line to your genetic algorithms, so that you arent to blame for the robot apocalypse
Alexfrog
May 6, 2009 at 12:52 pm