Tetris—: Exploring Human Performance via Cross Entropy Reinforcement Learning Models


What can a machine learning simulation tell us about human performance in complex, real-time tasks like Tetris—? Although often used as a research tool (Mayer, 2014), the strategies and methods used by Tetris players have seldom been an explicit focus. In Study 1, we use cross-entropy reinforcement learning (Thiery & Scherrer, 2009) to explore (a) the utility of high-level strategies for maximizing performance and (b) a variety of features and feature-weights (methods) for optimizing a low-level, one-zoid optimization strategy. Two strategies quickly rise to performance plateaus, whereas two others continued towards higher but more variable performance. In Study 2, we compare zoid (i.e., Tetris piece) placements made by our best models with those made by the full spectrum of novice-to-expert human Tetris players. Across 370,131 episodes collected from 67 human players, the ability of two strategies to classify human zoid placements varied from 43% for novices to 65% for experts.

Back to Table of Contents