No abstract

No abstract

No abstract

First and foremost, I am grateful to my three supervisors, Christina, Jakob, and Stephen, who have provided me with many lessons on both academia and science as a whole, as well as personal lessons I will keep forever. My weekly meetings with Christina and Jakob have always been a source of joy, as they were both scientifically enlightening but also simply entertaining. I would also like to thank my fellow PhD students, with whom I have worked during the years, Lucas, Dongsheng, Stephan, and Niklas.During my PhD, I have worked together with great people at Edulab to improve how mathematics is taught in Denmark. This has been an excellent collaboration where real-world problems could meet academia with exciting results. Of all the people at Edulab, I would especially like to thank Lotte, Kasper, Morten, Jakob, Stani, and Daniel.I was fortunate to spend a summer away from Denmark as an intern at Spotify Research in London. I got to work with many wonderful people, where I would especially like to thank Rishabh, Brian, Federico, Lucas, and Mounia.I also want to thank Rastin and Benjamin for a great collaboration at Danmarks Nationalbank. I realized how lucky I am that papers in computer science are so much shorter than in economics.From my personal life, I would first like to thank all my friends and family who have put up with my wacky work schedule during the last few years. Especially I would like to thank Mikkel, Helena, and David, who have been a great support. Lastly, I would like to thank the two most important people during my PhD, my twin brother Casper and my fiancée Spring. Casper has been a great source of inspiration in our work together and has been the best collaborator anyone could ever hope to work with. Spring has been a constant source of support and love, who not only accepts that I like to talk about my work, but also perfectly feigns interest in it.

Inspired by applications in sports where the skill of players or teams competing against each other varies over time, we propose a probabilistic model of pairwise-comparison outcomes that can capture a wide range of time dynamics. We achieve this by replacing the static parameters of a class of popular pairwise-comparison models by continuous-time Gaussian processes; the covariance function of these processes enables expressive dynamics. We develop an efficient inference algorithm that computes an approximate Bayesian posterior distribution. Despite the flexbility of our model, our inference algorithm requires only a few linear-time iterations over the data and can take advantage of modern multiprocessor computer architectures. We apply our model to several historical databases of sports outcomes and find that our approach a) outperforms competing approaches in terms of predictive performance, b) scales to millions of observations, and c) generates compelling visualizations that help in understanding and interpreting the data.