As the primary striker for such a juggernaut, it can be hard to detangle Messi's goal-scoring prowess from Barcelona's general offensive dominance. (...) I think this criticism is fair -- and I found it intriguing enough to look into the matter myself.
(...)
By now I've studied nearly every aspect of Messi's game, down to a touch-by-touch level: his shooting and scoring production; where he shoots from; how often he sets up his own shots; what kind of kicks he uses to make those shots; his ability to take on defenders; how accurate his passes are; the kind of passes he makes; how often he creates scoring chances; how often those chances lead to goals; even how his defensive playmaking compares to other high-volume shooters.
Yet none of that really addresses the main potential confounder the author acknowledges: most measurable aspects of a player's game --not only (assisted) goals but also even touch-by-touch, more seemingly 'individual' skill measures such as passing success-- are arguably heavily conditioned on the ability of the rest of the team. (Today, for instance, Messi didn't score against Switzerland, and the team was goalless for 117 mins until a tepid 1-0 in extra time (edit: but assisted by Messi! my apologies)... go figure).
A 'matched' design would have been much more appropriate, but I realize it's hard to do sound data analysis on a journalist's deadline.
I found this curious as well. Hockey has by far the most rudimentary statistics out of the 4 major North American sports plus Soccer and yet even the hockey stats community has the ability to show how well a team does with a given player on the ice vs. without.
Excellent point, but would you care to explain how you would match subjects in this case? They should be comparable pre-something, but what would that something be?
Perhaps a superficially plausible ignorability assumption would be that, during rapid breakouts from the defensive zone, striker skill would strongly depend on the midfield's ability to open up favorably, which in turn would depend somehow on the defense's overall (relative) skill; so that, viad-separation, striker skill and defense's relative skill would be independent conditional on midfield's skill during breakouts. It would definitely not be trivial!
(...)
By now I've studied nearly every aspect of Messi's game, down to a touch-by-touch level: his shooting and scoring production; where he shoots from; how often he sets up his own shots; what kind of kicks he uses to make those shots; his ability to take on defenders; how accurate his passes are; the kind of passes he makes; how often he creates scoring chances; how often those chances lead to goals; even how his defensive playmaking compares to other high-volume shooters.
Yet none of that really addresses the main potential confounder the author acknowledges: most measurable aspects of a player's game --not only (assisted) goals but also even touch-by-touch, more seemingly 'individual' skill measures such as passing success-- are arguably heavily conditioned on the ability of the rest of the team. (Today, for instance, Messi didn't score against Switzerland, and the team was goalless for 117 mins until a tepid 1-0 in extra time (edit: but assisted by Messi! my apologies)... go figure).
A 'matched' design would have been much more appropriate, but I realize it's hard to do sound data analysis on a journalist's deadline.