Models that play "forever" adopt different strategies than those that "die" fast.

“

Slow and steady wins the race, but the best performers may be those who know when to change their strategies. The differences in scores between our best tortoise and best hare model players (MPs) may be the most dramatic outcome of this work, but it is far from the most important. We see our contribution in terms of six bullets which we list in order of their INCREASING RELEVANCE TO HUMAN BEHAVIOR; namely, 1. the ability of small changes in the test environment (i.e., different random seeds for generating sequences of 7 numbers) to produce large changes in performance, 2. the failure to find one model (i.e., one set of six feature weights) that could override the importance of random number generation on zoid sequence (see AppendixA1), 3. the emergence of two different strategies for sequential decision-making, endurance for the tortoise models and escalation for the hares, and the consistency of those strategies across multiple environments, 4. the identification of endurance and escalation strategies in human behavior, 5. the finding that the endurance strategy is favored by less experienced HPs and the escalation strategy favored by our most experienced HPs (see our Table 9), and 6. the additional finding that even our very best human players switch to an endurance strategy when they reach the limits of their expertise (see our Fig. 3)

”