Predicting and Understanding Initial Play
-Micro Theory Lunch
PCPSE Room 202
*joint with Drew Fudenberg
Abstract: We use machine learning to uncover regularities in the initial play of matrix games. We first train a prediction algorithm on data from past experiments. Examining the games where our algorithm predicts correctly, but existing models don’t, leads us to add a parameter to the level-1 model that significantly improves predictions. We then generate new games where our modified level-1 model performs poorly, and obtain better predictions with a hybrid model that uses a decision tree to decide game-by-game which rule to use for making predictions. Finally, we show how to further improve predictions using crowd-sourced predictions as an input.