1. Math guys look into something called PECOTA - it's a sabermetric analytical tool for predicting the performance of baseball players by comparing them to previous players and generating similarity scores - i.e. predict the trajectory of a player's career by their existing data. It strikes me that this is quite similar to what we want to do.
If the game could analyze data and figure out which initial non-primary factors are most highly correlated with future success, we could generate expected development curves for each player which predicts their behavior. This is a semi-solution to the new player problem.
2. As the semester gets started again it would be nice to have a regular meeting data/time - please post availability.
3. An idea related to initialization - presumably the game designers know how their game is supposed to play, so what if we used their data as the seed? i.e. attempt to have the system build teams which play similarly to how dev teams played.
4. I believe windowing/historical weighting still isn't implemented, but it really needs to be.
5. I'm going to generate a document summarizing the current situation in the computer games industry, specifically trends in PC multiplayer gaming. I'll try to have something done by Sunday.
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment