We explored the behavioral game between a predator, the little egret (Egretta garzetta), and a prey, the common goldfish (Carassius auratus), in a laboratory theater containing three fish pools. We tested the hypotheses that the egrets maximize their total capture success by responding to the fish’s antipredatory behavior and that the behaviors of both players respond adaptively to the density distribution of fish among the pools. One experiment presented egrets with 15 fish per pool. The second experiment used a heterogeneous environment: pools 1, 2, and 3 had 10, 15, and 20 fish, respectively. Within each pool, fish could move between a safe, covered microhabitat and a risky, open microhabitat. Only the risky habitat had food, so fish were trading off food and safety by allocating the time spent in the two habitats. Egrets spent more total time in pools with more fish and returned to them sooner. Egrets maximized the number of fish they captured by following the matching rule of the ideal free distribution. The fish used the risky but productive habitat 65% of the time during experiments without egrets, but only 9% during experiments with 15 fish and egrets present somewhere in the theater. In addition, with egrets present, fish fine-tuned their behavior by reducing their use of the risky habitat as the egrets increased the frequency of their visits.