About

Me

My name is Martin Ingram. I’m a PhD Student at the University of Melbourne, Australia, supervised by Nick Golding and Damjan Vukcevic. I completed my undergraduate degree in Natural Sciences (Physical) at the Unversity of Cambridge in 2014 and a Master’s degree in Computing Science at Imperial College, supervised by William Knottenbelt in 2015. You can find my full resume here.

I’m interested applying Bayesian statistics to practical problems. You can view my publications on Google scholar.

My github account is here: github

One current focus is on researching variational inference algorithms that are both fast and accurate enough. I am doing some work on this using the JAX package in python. Some code for this is in the SVGP repository.

Papers

How to extend Elo: a Bayesian perspective Ingram (2021): Published in JQAS. This paper discusses how the Elo rating system, originally developed for chess but now popular in many sports, can be related to Bayesian inference and (extended) Kalman filtering. The connections are used to develop principled extensions of Elo which are shown to outperform existing methods when used for prediction.

Multi-output Gaussian processes for species distribution modelling Ingram, Vukcevic, Golding (2020). Published in Methods in Ecology and Evolution. We explain how multi-output Gaussian processes (MOGPs) can be used to model the distribution of multiple species. We use variational inference to scale the approach to large datasets and show that MOGPs outperform other approaches, especially for species that are rarely observed.

Space-Time VON CRAMM: Evaluating Decision-Making in Tennis with Variational generatiON of Complete Resolution Arcs via Mixture Modeling Kovalchik, Ingram, Weeratunga, Goncu (2020): Available on arXiv. We use an infinite Bayesian Gaussian mixture model to model player and ball trajectories in tennis and use it to develop a number of metrics, including an expected shot value, to evaluate players’ decision making.

A point-based Bayesian hierarchical model to predict the outcome of tennis matches Ingram (2019): Published in JQAS. I present a Bayesian hierarchical model to predict tennis matches, incoporating surface and tournament effects into a dynamic Bradley Terry model that is fit using Stan. PDF.

Estimating the duration of professional tennis matches for varying formats Kovalchik, Ingram (2018): Published in JQAS.

Adjusting bookmaker’s odds to allow for overround Clarke, Kovalchik, Ingram (2017): Published in the American Journal of Sports Science

Hot heads, cool heads, and tacticians: Measuring the mental game in tennis Kovalchik, Ingram (2016): Presented as a poster at the 2016 MIT Sloan Sports Analytics Conference. PDF.

Contact me

martin.ingram@gmail.com