Match-fixing is a key problem facing many sports, undermining the integrity and
sporting spectacle of events, ruining players’ careers and enabling the criminals behind the fixes to funnel funds into other illicit activities. Although for a long time authorities were reticent to act, more and more sports bodies and betting companies are now taking steps to tackle the issue, though much remains to be done. Tennis in particular has faced past criticism for its approach to combatting match-fixing, culminating in widespread media coverage of a leak of match-fixing related documents in 2016, although the Tennis Integrity Unit has since intensified its efforts to deal with the problem.

In this thesis, we develop new statistical methods for identifying tennis matches
in which suspicious betting activity occurs. We also make some advancements on
existing sports models to enable us to better analyse tennis matches to detect this corrupt activity. Our work is among the first to use both pre-match and in-play odds data to investigate match-fixing, and to also integrate betting volumes. Our pre-match odds are sampled at several intervals during the pre-match market, allowing for more detailed analysis than other work. Our in-play odds data are recorded during every game break along with live scores so that we can explore how the odds vary as the score progresses. In particular, we look for divergences between market odds and predictions coming both from sports models and from direct predictions of odds based on in-play events. Our methods successfully identify past matches that other external sources have found to contain suspicious betting activity, and are able to quantify how unusual this activity was in relation to typical betting behaviour. This suggests that our methods, coupled with other sources of evidence, can provide a valuable quantification of suspicious betting activity in future matches.