OptaPro Forum 2018 (slides)
Update (2018-10-12): A proper write-up of this presentation has now been published on the OptaPro website. You can find it, the slides, and the original submission here:
https://www.optasportspro.com/about/optapro-blog/posts/2018/blog-counting-across-borders/
A couple of people have asked for a copy of the slides to my 2018 OptaPro forum talk, so I’ve put them here, in case they’re of any interest to anyone else.
I suspect this example won’t age too well this season, but there are a couple of caveats that I mentioned in the talk. First off, the model was trained on the seasons up to 2015/16, at which point Aubameyang had just had his first 20+ goal season. Secondly, Aubameyang has played most of his career out wide, and since position isn’t included in the model (see the appendix), the model won’t credit his underlying goalscoring ability as much as it perhaps should in these seasons.
I should probably have gone with the Batshuayi example I initially had in there, but the temptation to crack a “he’s 28 until he’s 29” joke was too great 🤷♂️.
Soon after this, I learned that Paddy McCourt holds the nickname “The Derry Pelé”. Good to see the stats picking this up.
The appendix slides were an attemt to anticipate some questions I though I might get. This first slide shows that the rate at which players complete take-ons decreases slower than the rate at which players attempt take-ons.
I believe this shows that players become more selective about when they attempt a take-on as they age, as well as attempting fewer overall.
This shows an earlier attempt to test the inclusion of positions in the model. I didn’t like the results, though. I think the main problem was that the position data used wasn’t granular enough (lots of roles came under the same “position”, for example), and that players don’t change position enough for the model to accurately differentiate player and position effects.
There are of course other issues with including position like this such as the effect of position on goalscoring changing depending on the player themselves.