23 April 2017. El Clásico. Real were one man down and losing against one of the best teams in the world, with less than 15 minutes to go. The match seemed decided. But then, Real equalized. Looking at the usually tracked events will paint a rather boring picture of the goal: Kroos passes to Marcelo, who crosses to James, who scores. But how did James get in such a great position to shoot from, and who allowed it? We can`t say. Looking at the video replay will give you much more insight. The main reasons Real got the opportunity to score were: Barcelona were quite relaxed defensively, allowing equal number of attackers and defenders Ronaldo`s movement opened up space in the near post James got rid of the marking with a great run into the near post Marcelo crossed perfectly towards James` run, allowing him to shoot with no opposition A lot of people bring examples like this one to show why stats alone can`t describe what happens in a football field. Indeed, you can`t objectively measure these 4 elements just by looking at events data. However, I will argue that you can measure them if you have full positional data (i.e, coordinates for every player on the pitch at every moment of the play), creating relevant stats that measure how space is created, used — and also denied. How do we improve this metric? Incorporate the improved passing options definitions (specifically passing difficulty) Consider potential changes in the recipient movement and available space, converting the destination spots into destination zones Why do we need this metric? We have really good information about passes made— origin and destination coordinates, type of pass, success rate, and so on. However, we don`t really measure if that pass was the better option at the moment it was made, and we don`t really measure if the pass was made to the right place at the right speed. We also don`t know how difficult was that pass: was the receiver completely free? Was there a defender blocking the passing line? Great passers risk a lot, especially up front, resulting in a lower than normal success rate. Measures to correct this usually involve considering the origin and destination of the pass, but available space should be the most important factor to evaluate passing difficulty. Conclusion It is not a coincidence that advanced sports analytics started in Baseball: the sport lends itself to easy record-keeping, with constrained individual duels (the pitcher vs the batter) as the basis for everything that happens. Football, in contrast, is a hard sport to summarize: it mixes the free-flowing nature of Basketball with the tactical complexity of American Football. Unlike Basketball, possession chains are murky in football (with lots of duels and no clear owner of the ball). Unlike its American cousin, you can`t train a fixed playbook (outside of Set Pieces, football is… well, messy) and there`s no quarterback dictating play. The goal of this article was to convince you that we need fully positional data to properly measure what happens in the field. Although the top tier professional teams are already doing this type of analysis, we, the fans, don`t have access to it. But we need television broadcasters, sports journalists and tournament organizers to provide such value-added analysis. Hopefully, this post will at least spark interest in it.
© Copyright 2017 Alle Rechte vorbehalten.