As an industry first, state-of-the-art statistics are being applied to rare events and ground subsidence due to underground coal mining. Results are being used to improve the accuracy of recommendations made to the Government and mining companies. ISG developed these statistical methods using Extreme Value Theory (EVT) in order to improve prediction of the magnitude of ground subsidence, and the consequential impacts on structures. The first stage of this work involved exploratory statistical analyses to quantify the probability that a future ground strain caused by mining exceeds a specified maximum tolerable subsidence (that is, a trigger point). It was demonstrated that using EVT motivated models to describe the extreme tails of observed strain data resulted in more credible fits than those based on alternative models originating from the full dataset. As a consequence, the predictions of future extreme subsidence in excess of the trigger points are more reliable. In addition, the use of regression methods improved accuracy and precision by the inclusion of relevant explanatory variables, such as the distance from the point of interest to the mine (a ‘far field’ analysis), and the modelling of the relationship of subsidence strain and curvature. Work on the project incorporates the effect of multiple ‘longwalls’ (the mines excavated by drilling equipment moving underground), as well as smoothing raw curvature data. The application of statistical EVT to predict ground subsidence is a new application of the theory; it is also the first time that state-of-the-art statistics has been used in this particular industry sector.