THREAD: Using VAEP data metrics to highlight England's most valuable players at the 2018 World Cup.

Football almost came home... but who contributed the most? Let's find out.

#England #WorldCup #ItsComingHome #Python #DataScience
First of all, what is VAEP?

VAEP (Valuing Actions by Estimating Probabilities) is a metric that evaluates a player’s offensive and defensive contributions to their team. This is achieved by calculating expected value gained/lost by the player’s actions - passes, dribbles, etc.
Offensive contribution is measured by how much a player's actions with the ball (e.g. passes, dribbles) improve their team's chance of scoring.

Defensive contribution is measured by how much a player's actions limits an opponent's chance of scoring.
Below is a very good example of VAEP in action, courtesy of @GoalAnalysis - he has already produced a lot of work in this area so make sure to check it out!
First, a look at how Southgate's men contributed to their team's scoring chances.

Two CBs and a set-piece taker (Trippier) rank in the top five, highlighting England's reliance on attacking set pieces throughout the tournament.
Next, a look at raw values for defensive contributions.

John Stones stands out as England's top contributor towards limiting opponents. Loftus-Cheek and Alli also appear to have put in contributions from the middle of the park. The rest of England's midfield? Not so good...
Finally, a look at England players' contributions per minute played. Circle size represents minutes played.

Stones was an underrated member of Southgate's side, putting in work at both ends of the pitch.

As for Eric Dier, he'll always have THAT penalty...
A limitation to this analysis is that the model used to compute VAEP does not appear to take into account off-the-ball movements in attack.

Raheem Sterling excelled at creating space with his runs throughout the tournament, but this was overshadowed by lack of output.
For further reading into VAEP, I strongly encourage reading the paper published by the team behind the metric: https://arxiv.org/pdf/1802.07127.pdf

I was able to obtain this data by utilising the SoccerAction Python package, which can be found here: https://github.com/ML-KULeuven/socceraction
That just about wraps it up! Would appreciate any comments/feedback, and I hope to make more threads like this in the future!

All data taken from @StatsBomb's Open Data repository: https://github.com/statsbomb/open-data
You can follow @CharlieNData.
Tip: mention @twtextapp on a Twitter thread with the keyword “unroll” to get a link to it.

Latest Threads Unrolled:

By continuing to use the site, you are consenting to the use of cookies as explained in our Cookie Policy to improve your experience.