Analysis of upset matches in Overwatch League 2019 — Methodology

Firstly I’ll define what I mean by an upset match. Simply put an upset match is when the actual outcome of the match differs from the predicted outcome.

To determine whether a result is an upset or not, I considered whether the winner was a higher seed than the loser, based on the result of the previous stage. This is why I haven’t drawn any conclusions from stage 1, as there was no previous stage to use to determine the seeding.

I could have done something based on the previous season, but as this wouldn’t have given us any data for expansion teams it seemed better to not attempt this rather than to have incomplete data.

Another area of my analysis that warrants explanation is what I’ve called the ‘upset factor’. The upset factor is my attempt to determine mathematically how big an upset is. This is achieved by calculating the difference between the seeding of the teams. For example the San Francisco Shock (seeded 6th in stage 1) beating the Vancouver Titans (seeded 1st) in the Stage 2 final would have an upset factor of 5, because they beat a team that was 5 seeds higher according to our best estimates.

I’ve also tried to estimate whether an upset is a major or a minor upset. The criteria for determining this was whether that gap between the seeds was greater than the average for upset matches plus 20% and rounded up for ease of understanding.