In this paper we ask a fundamental design question in the theory of contests: Should contestants and reward money be pooled into a single grand contest or should they be divided into parallel subcontests? We theoretically explore optimal divisioning using Tullock’s lottery contest framework and compare the performance i.e. total effort generated of the grand contest to the performance of contest divisioning based on players ability or risk attitude. When all players are risk neutral contest divisioning is never optimal. However given that players are heterogeneous in either ability or risk attitude we find that contest divisioning is optimal when the degrees of absolute risk aversion are sufficiently large. Importantly our results support the widespread use of divisioning seen in practice.