We study sequential parimutuel betting markets with asymmetrically informed bettors, using an experimental approach. In one treatment, groups of eight participants play twenty repetitions of a sequential betting game. The second treatment is identical, except that bettors are observed by other participants who assess the winning probabilities of each potential outcome. In the third treatment, the same individuals make bets and assess the winning probabilities of the outcomes. A favorite-longshot bias is observed in the first and second treatments, but does not exist in the third treatment. Information aggregation is better in the third than in the other two treatments, and contrarian betting is almost completely eliminated by the belief elicitation procedure. Making bets improves the accuracy of stated beliefs. We propose a theoretical model, the Adaptive Model, to describe individual behavior and we find that it effectively explains betting decisions, especially in the third treatment.