The big problem for me with looking at the numbers in isolation is that they don't really tell the whole story.
I don't want to appear patronizing or facetious here, but surely no-one would claim that they do? Nobody just checks a BBC data feed on possession, shots, corners, fouls, etc. instead of watching the game, and then assumes that they can recreate the game in their mind's eye, do they?
Stats are like maps. A map never captures all the diversity of a landscape, and stats will never capture the complexity of a game (especially a game that isn't broken up into discrete segments like football). But they have a variety of uses: they can allow you to investigate long-term trends, they can provide an objective measure of a team's success in one facet of the game, they can challenge your subjective opinions about an individual's patterns of play. They're just tools, and like any tool, they can be misused. So, I'd suggest that a reasoned response to a statistical argument shouldn't be based on the contingencies of a particular game, but whether a particular set of data, or an analytical framework, is the best way to summarize a process or an outcome.
So, for example, is "shots ratio" the best index we have of how a team has been performing, or how it will perform in future?
