Appendix 1—table 5. Comparison of prediction accuracy (± SE) across four models in four game contexts for Monkey O.

	Strategy
Context	Dynamic	Static	LARL	Perceptron
Overall	0.907 ± 0.008	0.806 ± 0.014	0.659 ± 0.016	0.632 ± 0.013
Early game	0.868 ± 0.0132	0.772 ± 0.054	0.765 ± 0.042	0.548 ± 0.037
Late game	0.901 ± 0.016	0.786 ± 0.027	0.595 ± 0.026	0.598 ± 0.026
Scared ghosts	0.952 ± 0.019	0.729 ± 0.047	0.658 ± 0.046	0.545 ± 0.050

LARL: linear approximate reinforcement learning.