The prices offered by the fixed-odds bookmakers in the tennis betting market are biased because of the favorite-longshot phenomenon. How to derive unbiased implied probabilities underlying the published odds is the focus of this study. This paper proposes a new normalization procedure that yields unbiased probabilities, regardless of the presence of heavy underdogs. In-sample, the proposed normalization has a superior forecasting ability than the other methods. Moreover, it enables betting strategies, which produce superior returns than those obtained from the Bradley Terry type model.