Model Solutions In Tennis Prediction

Statistical modelling is nowadays found in more spheres of life than ever before, and sport is no exception. This can be seen every week in World Ranking systems, or in the Expected Goals stats on Match Of The Day, but is also being used behind the scenes in training and transfers, not to mention integrity commissions, who might try to use market information to help identify those trying to fix matches. All of these commercial applications require an understanding of predictability – how do we tell from what we’ve already seen what will happen in the future? Such questions are challenging, and answering them requires mathematical & statistical thinking – and a healthy knowledge of sport…

In this talk, Dr Nelson will discuss his own personal experiences that led to a career at ATASS sports, a world-leading sports forecasting firm, before discussing exactly what being in sports forecasting involves. We’ll discuss both current statistical methods and tools from machine learning, and the pitfalls to be avoided when using them, all illustrated by the following problem: how do we predict the probability of a player winning a tennis match from world rankings alone?