This paper analyzes repeated procurement of services as a four-stage game divided into two periods. In each period there is (1) a contest stage à la Tullock in which the principal selects an agent and (2) a service stage in which the selected agent provides a service. Since this service effort is non-verifiable, the principal faces a moral hazard problem at the service stages. This work considers how the principal should design the period-two contest to mitigate the moral hazard problem in the period-one service stage and to maximize total service and contest efforts. It is shown that the principal must take account of the agent's past service effort in the period-two contest success function. The results indicate that the optimal way to introduce this `bias' is to choose a certain degree of complementarity between past service and current contest efforts. This result shows that contests with `additive bias' (`multiplicative bias') are optimal in incentive problems when effort cost is low (high). Furthermore, it is shown that the severity of the moral hazard problem increases with the cost of service effort (compared to the cost of contest effort) and the number of agents. Finally, the results are extended to more general contest success functions. JEL classification: C72; D82 Key words: Biased contests; Moral Hazard; Repeated Game; Incentives.