What this blog is about

Welcome to anyone who happens upon this blog!

This blog is meant to document my amateur attempts to build a backgammon player. Initially my focus is on building a neural net player akin to TD gammon.

I've actually tried three times to get a network up and learning properly. The first two times were failures - I couldn't find the bug(s), but they never actually learned to play. The third time shows some promise, and this blog is meant to document how I approached the problem and how my approach worked. Once I've got the basic machinery working I'll continue to extend it, and document that.

I'll also post interesting links or research I stumble across. As I said, I'm just an amateur, so I'm probably ignorant not just of state of the art in computational backgammon, but of backgammon itself. I'm a decent intermediate player but certainly no expert. I just find the numerical machinery behind artificial learning really intriguing and this is a way for me to play with it.