Phylourny: efficiently calculating elimination tournament win probabilities

Phylourny: efficiently calculating elimination tournament win probabilities via phylogenetic methods

When predicting the outcome of knockout tournaments, such as those typical in football tournaments, the traditional method is to fit a model using historical data, and then use that model to simulate a tournament using a Monte Carlo method to obtain predicted win probabilities. However, using lessons and algorithms from computational phylogenetics, we can instead compute the win probabilities for a knockout tournament exactly for some specific model while also computing the results significantly faster (2 - 3 orders of magnitude). We implemented these techniques into a tool called Phylourny.

In addition, we use Phylourny to apply further techniques from computational phylogenetics. Specifically, we explore the parameter space for a given model using a Monte Carlo Markov Chain and summarize the stability of results for those model parameters. We show, by example with the several tournaments, that exploring the parameter space produces a more robust predictions, as well as characterizes the confidence of the prediction given.

The talk was given by Ben Bettisworth from the Heidelberg Institute for Theoretical Studies