I think it has some good points but I’d be wary of taking his result from the regression without question. Linearity may be the way to go but I’d be cautious of it not capturing the issue with stat padding in certain games and not capturing the idea that defences will lead to fewer dropped points (particularly against rivals). Indeed, more I think about it, the more I think linearity may be a poor model. I also feel like using City (who is something of an outlier across the whole PL) may lead to some issues, and the initial opening paragraphs are overly simplistic. Point is, algorithms will always give you an output, doesn’t mean the algorithm is suitable. I’m not disputing the result (I generally don’t put too much stock in any ‘truism’), and think it raises valid points, but wouldn’t say it is a clear slam dunk. I think it may also be worth exploring the points taken to win the league and seeing how that effects things, i.e. if points taken < threshold then better defence is preferable, above then better attack, that kind of thing.