Statsbomb/FBref kinda do this with 'Goal Creating Actions'.

The problem still is Basketball is high scoring so it's relevant.

Football's problem is variance. I mentioned before that there is 2.5 goals per game. So in a match that produces ~3500 lines of event data, 0.07% will https://twitter.com/OneandonlyAK/status/1387192376171638787

be for goals.

In basketball there are maybe 50 to 60 scoring events per game.

I used Mane and Milner as examples a few seasons back. Milner 7 assists. Mané 1. Mane had created more big chances, had a higher xA but ultimately what he created wasn't converted, Milners was.

Rewind to the season before, Mane was 7 assists. He hadn't massively changed his game it's just that the chances he created were converted. He actually had an almost identical xA (5.25 in 17/18 and 5.12 in 18/19).

And I think we get misguided a LOT by variance jumping to big

conclusions. Another example is Greenwood.

I suggested last year that overperforming xG to the point where his finishing was ~10 times better than Messi wasn't sustainable. That it was more likely a purple patch on limited data rather than 'this is how his career will be'.

And then start of this season we see the opposite where he gets 1 goal in ~23 games.

Even now we have too little data to really assess Greenwood. He only has about 3000 minutes of football which is ~33 full games of data. And a lot of that is cameo appearances which can mislead.