Forecasting with Colonel Blotto

In this paper we design and test a competitive forecasting mechanism based on the Colonel Blotto game. In the game, forecasters allocate a fixed number of resources to different 'battlefields'. Each field is realized with a probability that is determined by a stochastic process. Subjects learn about the underlying process during the course of the experiment and thereby form beliefs about the probability that a field is selected. Once a field is selected, the subject competes for a payoff that is associated with the number of resources allocated to that field. We implement two different payment rules, a lottery and an auction, and find that the lottery outperforms the auction. This relative underperformance of the auction can be attributed to the strategic uncertainty being too high in the auction and the strong incentives to misalign allocations.