Analytics in Football - Official Thread

I was wondering about that graphic, I was surprised Messi was so high, then I noticed the x-axis is raw goals. In my opinion, if it was like a percentage it would make more sense.

Maybe. The risk, if you don't put in an "xG of < threshold > or more" filter, is that you'd get a lot of noise. Like centre-backs who only get the odd headed shot on target who happened to get get luck with 2 or 3 across a season.

I'm still kind of surprised he's so high because I'd have thought his free kicks would have dragged him down a bit.

Anecdotally it feels like he's bloody good at free-kicks and often works the keeper. When compared with the average free-kick taker, he's still probably exceptionally good.

CR7, by contrast, has had an awful free-kick record for years but still insists on taking them.

But maybe I just don't know how it's 'calculated'. Is it based on where the scorer receives the ball initially? I thought until now it was based on the shooting.

xG is based largely on where you took the shot on. If you beat 4 guys before taking the shot you won't get any extra xG credit for that. Increasingly xG is starting to consider other factors such as positioning of opposition players when you took on the shot, height that you made contact at etc.

I'm still kind of unconvinced about the value of the metric as giving much meaningful information.

Also, what's the y axis in that visualization? They should have used seasons as it would make it easier to track progress, but I think they just spaced it out at random.

Good question. It's not clear. It's definitely not shot volume, amount of goals or xG per 90.

Feels meaningful that all of the prominent under & overperming seasons are tightly bunched in a row, but what is the meaning? ;)
 
Maybe. The risk, if you don't put in an "xG of < threshold > or more" filter, is that you'd get a lot of noise. Like centre-backs who only get the odd headed shot on target who happened to get get luck with 2 or 3 across a season.



Anecdotally it feels like he's bloody good at free-kicks and often works the keeper. When compared with the average free-kick taker, he's still probably exceptionally good.

CR7, by contrast, has had an awful free-kick record for years but still insists on taking them.



xG is based largely on where you took the shot on. If you beat 4 guys before taking the shot you won't get any extra xG credit for that. Increasingly xG is starting to consider other factors such as positioning of opposition players when you took on the shot, height that you made contact at etc.



Good question. It's not clear. It's definitely not shot volume, amount of goals or xG per 90.

Feels meaningful that all of the prominent under & overperming seasons are tightly bunched in a row, but what is the meaning? ;)
My quibble wasn't so much the threshold but using raw goals amplifies noise. Yes, it should revert to the mean, but if you think of each chance as an event in a stochastic process, like a random walk, the more chances a player gets, the more the variance widens.

Ronaldo has indeed a lower conversion rate than Messi, and apparently he's gotten worse over the years. Messi improved over time, as he did at many other aspects of his game, but back then I'd have thought he was fairly inaccurate. But maybe a free kick outside the box is less than 5% chance of a goal anyway.

Maybe Messi outperformed because he usually was one on one against the keeper from inside the box, whereas most players would have defenders blocking from the same position?
 
My quibble wasn't so much the threshold but using raw goals amplifies noise. Yes, it should revert to the mean, but if you think of each chance as an event in a stochastic process, like a random walk, the more chances a player gets, the more the variance widens.

Yep, I take your point there. I think he probably chose the raw goals approach as the narrative he was searching for was probably:

"Who are the players who hugely over/underperform xG...and are heavily in the mix in terms of having chances to score"

Ronaldo has indeed a lower conversion rate than Messi, and apparently he's gotten worse over the years. Messi improved over time, as he did at many other aspects of his game, but back then I'd have thought he was fairly inaccurate. But maybe a free kick outside the box is less than 5% chance of a goal anyway.

I don't ever recall Messi being poor at free kicks. His style of striking the ball is very controlled and tends to go for accuracy over pace.

As you say, the xG on most free-kicks isn't going to be huge - so any underperformance of xG is unlikely to hurt a players xG efficiency much overall. On the flipside, being an exceptional free-kick taker might be a good way to nudge a player into xG overperformance.

Cherry-picking a player who I expected might fit that mould: James Ward-Prowse. His career xG overperformance is pretty much entirely down to his free-kick prowess.

1667987993288.png

Maybe Messi outperformed because he usually was one on one against the keeper from inside the box, whereas most players would have defenders blocking from the same position?

That's not it. Those factors would be baked into the xG - i.e. if Messi really gets a higher % of 'easy' chances then he'd still need to convert them way more efficiently than the average player.

It's hard to be accurate on cause & effect for xG performance over such a large sample size, but some factors which may in some small part explain Messi's overperformance may be:

- He chooses really well when to shoot (or not)
- He strikes the ball very well. Tends to favour placement over power, but when needed he can put some fizz on it & still retain control. Struggling to remember him fluffing a shot badly
- More than your average player, he seems to have a good radar for what the keeper is up to (e.g. positioning, body language) etc. which feeds into his shot placement decisions
 
I had a play around with the Group stage stats data on FB Ref and produced a couple of visualisation using Python.


If you've got any interesting ideas around metrics to visualise ping them my way and I'll see if I can produce something decent.

HsxxM8Q.jpg

Not surprised to see Griezmann & Mbappe so high. I was surprised to see Alba in here but no Pedri - turns out they were just either side of the 5 SCA per 90 threshold. For Belgium this tournament probably marks the passing of the creative baton from Eden to Thorgan Hazard, with the younger brother racking up SCAs at twice the rate of his big brother.

I'll re-run this again at the end of the tournament and increment the minimum no. minutes to weight it more towards player who went deep in the competition.

Caveat: For the sake of visuals I manually omitted Lee Kangin of South Korea - who was off the charts with 18 shot-creating actions per 90 and passing accuracy of a shade under 80% across 126 mins of play.

RunzbTi.jpg
 

TheEarlofPearl

Full Member
Was there a book that focused on analytics at Spurs in recent years. Heard someone mention it but cant seem to find it on amazon. It might be this but it doesn’t mention Spurs … https://www.amazon.co.uk/gp/aw/d/1788702050/ref=ox_sc_act_image_1?smid=A3P5ROKL5A1OLE&psc=1
This one?
 

 
H2XVyEf.png

Worked on this graphic last night. Some familiar/expected faces (Koulibaly, Timber, Hakimi) and a few less familiar ones (who were typically under the cosh a fair bit during the group stage).
 

EVENT GUIDE - HIGHLIGHT
Columbia Mills
Coughlan's, Douglas St.

30th Jun 2023 @ 7:30 pm
More info..

Rashomon

Triskel Arts Centre, Tomorrow @ 6pm

View more events ▼
Top