This paper examines the theoretical background and actual behavior in a gam
ing tournament with endogenous timing where a person has more incentive, st
ructure, and time to form a strategy. The baseline treatment suggests that
subgame perfection is a reasonable predictor of behavior - subjects made 17
0 of 208 theoretically predicted choices of best actions, with the majority
of mistakes made in timing choices by the players who did not survive the
cut to the second round. Four sensitivity treatments established that the d
esign feature that lead to more predictable behavior was time to think - 74
5 of 960 correctly predicted decisions with more time versus 595 of 960 wit
h less time. A random effects Probit model suggests that the key design fea
ture that closed the gap between predicted and observed behavior was not ne
cessarily the non-linear payoffs created by the tournament design, but rath
er that the key was providing people with more time to think about their st
rategy.