Wednesday, November 06, 2002
( 9:00 AM ) Matt
The Jumpy Game
Okay, this isn't go related, but last night I found time to finish a pretty cool rendition of something my wife and I call "the Jumpy Game". It's given me some insight into how the front end for OkonomiGo might look and work.
On a more go related note, I am really excited about implementing some of the crawler style algorithms. I've decided to try most all of the meaningful values of the four constants, and put them up against each other. This should give some insight into setting up a ranking system. I wonder if there will be any strong rock-paper-scisors combinations.
Another very similar strategy to the one described yesterday is to evaluate each move as a function of the immediately adjacent 4 intersections. Each adjacent point can be white, black, empty, or an edge. So, there are 4^4 = 64 combinations. (Minus a couple if you don't want to deal with 1 by n boards.) A partial order of these 64 combinations describes a strategy. There are a whole lot of such partial orders, (something like 64! * 2^63). A brute force approach to examining this search space is not feasible. It would probably be best to either use a genetic algorithm to search it, or reduce it to a class of meaningful strategies. Some partial orders are obviously silly. For example, we would expect that there should be some symmetry to the strategy.
That's my rant for the day.# -
Comments: Post a Comment