The gambler's fallacy is defined as the avoidance of a winning outcome in a stochastic environment with a constant probability. We tested the possibility that the gambler's fallacy in humans is responsive to the amount of time between choice allocations. Two groups of subjects were placed in a six-choice betting game in which the choices were clustered into two "patches." Groups were defined by the length of time - 2s or 6s - between trials. On any given trial subjects allocated six points among the alternatives, and retained any points that were bet on the winning alternative. Both groups showed evidence of the gambler's fallacy bias. However, the bias was stronger in the 6-s ITI group than in the 2-s ITI group. This difference was found primarily to be due to differences in the number of subjects showing an opposing bias to the gambler's fallacy, namely a preference for the most recent winning alternative. This choice bias is termed the hot hand fallacy. Our findings contradict predictions derived from a foraging heuristic and from traditional accounts of the gambler's fallacy.