These are the show notes for Episode #10 of the Soccermetrics Podcast, with guest Jonathan Gruhl. This interview was recorded on 10 March 2014 from my Atlanta office.
Jonathan Gruhl is a fellow Stanford engineering graduate who has applied qualitative analysis to a variety of fields in his career, from management consulting to finance to sports. He currently works as a data scientist as a technology startup in the San Francisco area. Jonathan was one of the presenters at the OptaPro Analytics Forum and he gave a very interesting presentation on development of player similarities in football using in-match Opta data. There have been other works on player similarity in soccer before, and this wasn’t even the only such presentation at the conference (Marek Kwiatkowski’s work on similarity among midfielders was another), but Jonathan’s work starts us down the road toward the type of player projection models developed in other sports.
In this episode, Jonathan and I discussed the following:
- Similarity scores and why they’re challenging to develop in soccer
- Similarity of player outcomes vs similarity of player styles and actions
- Where Euclidean distance falls short as a way to judge similarity
- Who is similar to Yayá Touré?
- How unique are Landon Donovan and Gareth Bale?
- Is there much similarity between MLS and Premier League players?
Web links to relevant resources referenced during our conversation:
As always, we welcome your feedback on the episode, whether on content or technical matters. You can make your feedback in the comments section (recommended for content-related discussions) or by emailing [email protected] (recommended for technical issues).