Appendix 1—table 6. Comparison of prediction accuracy (± SE) across four models in four game contexts for Monkey P.

	Strategy
Context	Dynamic	Static	LARL	Perceptron
Overall	0.900 ± 0.009	0.825 ± 0.012	0.679 ± 0.015	0.615 ± 0.014
Early game	0.898 ± 0.017	0.824 ± 0.021	0.781 ± 0.022	0.605 ± 0.035
Late game	0.894 ± 0.016	0.826 ± 0.026	0.653 ± 0.025	0.599 ± 0.028
Scared ghosts	0.962 ± 0.011	0.723 ± 0.041	0.682 ± 0.027	0.456 ± 0.037

LARL: linear approximate reinforcement learning.