Ranking sports teams using probability forecasts from betting odds

Our research was motivated by the practical problem of rating and ranking European basketball teams on team strength. Typically, we would fit a latent strength rating model, such as the classical Elo rating, using observed outcomes. However, in most practical cases, this is insuficient for an accurate estimate of a team's strength, especially in the early parts of the season. We hypothesized that we can deal with this issue by taking into account probability forecasts derived from betting odds.

Initially, we based our approach on fitting the Elo model to probability forecasts . This proved to be a substantial improvement both in terms of higher predictive accuracy and fewer games needed for the teams' strengths to stabilize. Later, we adopted a Bayesian approach and extended the model to incorporate more detailed outcomes and probability forecasts. We'll look at the current state of this research, its potential application to detecting suspicious match outcomes, and some outstanding issues.