Update April 12, 2018: This post is from spring 2017, so the results reflect statistics up to season 2016–2017. I have now re-run the analysis including data from season 2017–2018, and updated results are available here. Methodology is unchanged, so … Continue reading →