This paper is about a Puyo Puyo chaining AI using a Deep Q Network. I think that's similar to what Google DeepMind uses? https://t.co/zVyR5nTt7H

This paper talks about using a Monte Carlo algorithm to play "incomplete information games", such as Puyo Puyo https://t.co/Socggeee01 Puyo is an incomplete information game because you only see two pieces ahead. To plan further ahead than that, an AI has to search a huge space.

Here's another paper about an AI algorithm that creates an 11.75 chain, on average: https://t.co/gRzcZB6EYs It's interesting how their abstract compares Puyo Puyo to Go and Shogi. I guess Puyo demands a similar kind of complex pattern recognition to determine the best move.

I need to study Japanese so I can understand these academic research papers on Puyo Puyo... Here's an interesting one. The authors worked on a way to procedurally generate Nazo Puyo puzzles, with a machine learning algorithm to rate its difficulty and fun https://t.co/y5MwWemXO9