We study the foundations for the development of optimal randomization in mixed strategy games. We consider a population of children and adolescents (7 to 16 years old) and study in the laboratory their behavior in a nonzero sum, hide‐and‐seek game with a unique interior mixed strategy equilibrium where each location has a known but different value. The vast majority of participants favor the high‐value location not only as seekers (as predicted by theory) but also as hiders (in contradiction with theory). The behavior is extremely similar across all ages, and also similar to that of the college students control adult group. We also study the use of cheap talk (potentially deceptive) messages in this game. Hiders are excessively truthful in the messages they send while seekers have a slight tendency to (correctly) believe hiders. In general, however, messages have a small impact on outcomes. The results point to a powerful (erroneous) heuristic thinking in two‐person randomization settings that does not get corrected, even partially, with age.

Developmental decision making laboratory experiment mixed strategy randomization C72 C93