For what it's worth, that's also what I do. But is it optimal? :) The source code is leading the blog posts [1], so I think I already have an answer to that qu

For what it's worth, that's also what I do. But is it optimal? :) The source code is leading the blog posts [1], so I think I already have an answer to that question for games on 2x2 and 3x3 boards using Markov Decision Processes, but the approach doesn't scale to the 4x4 board. [1] https://github.com/jdleesmiller/twenty48 https://github.com/jdleesmiller/twenty48