Headed home
Welcome to the 21st edition of Plot the Ball for 2023.
If you missed the previous edition, you can read it here:
Another brief, football-themed edition of the newsletter today — following on from Australia’s loss to England in the World Cup semi-final on Wednesday, and some of the discourse around their star striker Sam Kerr which has followed.
I’ll be taking part in a virtual event with The Data Lab Community on Friday 31st August, talking about the work I do in this newsletter. It’s free to attend, and you can sign up here.
Does Sam Kerr ‘miss easy chances’? It's complicated
In front of a home crowd of more than 70,000 people in Sydney on Wednesday, Sam Kerr scored one of the goals of her life.
In a heartbreaking end for those Matildas fans, though, England squeaked through a closely contested match with two late goals — and there will be no fairytale ending for Kerr and her teammates in the final on Sunday.
As well as her goal — from 32 yards out, according to FBref — Kerr also missed a couple of much higher-probability chances in the last 10 minutes of normal time.
Such is the life of a professional striker.
In the aftermath, Michael Cox of The Athletic — a thoughtful observer of women’s football — called back to an interesting comment he’d made about Kerr’s tendencies ahead of the tournament:
In Cox’s view, Kerr’s semi-final performance was the “ultimate example” of this.
But is such a reputation supported by the available data?
Kerr has now played six seasons1 of club football for which FBref (via Opta) provides expected goals data on its site — during which she has taken over 400 non-penalty shots2 and scored almost 100 goals.
When you start to group those shots into buckets based on the probability of scoring, though, the data gets a bit patchier — and we should temper the strength of our conclusions accordingly.
That said, there’s certainly no evidence that over this period she has missed shots she ‘should’ score3 — defined here as those with an expected-goal value of more than 0.50 — more often than the average striker in top-level women’s football.
She has taken 33 such shots in this stretch of her career, and scored 21 goals from 21.1 xG4.
Broaden things out a bit and look at shots with an expected-goal value between 0.30 and 0.50 — still incredibly good chances, as these things go — and it becomes clearer that there’s probably nothing to worry about.
She has scored 29 goals on 66 shots in this category, comfortably exceeding their aggregate xG value of 26.0.
Where it gets more interesting is on the other side of Cox’s assessment.
On the lowest probability shots — those with an expected-goal value of 0.10 or lower — we can be more certain that Kerr does in fact add a significant amount of value over the average striker.
She has taken 277 shots in this category, and outperformed expectation by more than 50% — scoring 25 goals on just 16.2 xG.
By normal standards, then, there is no reason to worry about Kerr’s ability to finish high-value chances — but Cox is correct that it’s her tendency to execute shots which have a high-degree of difficulty with relative ease that makes her truly superhuman5.
You can find the code for this piece on GitHub here
One season for Perth in Australia’s W-League, one for Chicago in the NWSL and four for Chelsea in the Women’s Super League.
It’s worth noting that FBref’s match shooting logs appear not to reconcile exactly with the aggregates on its individual player pages; if anyone knows why this is, please let me know.
This is trying to put a more analytical framing on a bit of terminology that is typically used by more traditional observers; ‘should score’ is understood in the sense that an average player is more likely to score a given chance than not.
You could certainly argue that, as a top striker, you’d expect Kerr to score more goals than her xG in these situations — and I’d probably agree. That said, I just don’t think there’s enough evidence yet to rule out this possibility going forward; if she were to score three of her first four 0.50 xG shots in the 2023-24 WSL season, for instance — an outcome that is perfectly plausible — she would be sitting at around +4% relative to her xG on chances in this range. When the number of shots you’re dealing with is this small, the metrics can be volatile.
If you want to read more about Kerr’s career — including those earlier seasons, for which advanced data is not available — I’d urge you to go and read ABC’s incredible piece of work on the Australian striker from late last year.