Interesting article, but there's so much within it that seems overstated.
The first thing to address is that we are dealing with probabilities here. Just because David Silva passes to Dzeko on the ground 50% of the time, doesn't mean that you should ignore the ball over the top to Balo. Statistical analysis is a help that allows us an unbiased evaluation of some aspects of the sport, rather than a prediction mechanism.
Secondly, and I say this as a computer programmer, there is no way at all to model football correctly programmatically. The reply that the article gives to this is:
For a start, good mathematicians can handle complex systems. At Chelsea, for instance, one of Forde’s statisticians has a past in insurance modelling. Football – a game of 22 men played on a limited field with set rules – is not of unparalleled complexity.
The problem though, is that this isn't a system of 22 played with set rules. This is a system of 22 variabe within a limited space and a set ruleset that
entirely depends on numerous variables for each decision taken by each of the 22 variables. This would be an NP-complete problem, it's the travelling salesman with 22 salesmen and an unlimited number of cities with each salesmen only allowed to be in one at a time. Computationally, it is impossible to model the game to any degree of accuracy which is why games like Football Manager don't try and just use a very simple decision making overlord system instead.
Football is a game of space; the creation, defending and exploitation of it.
Statistics are currently useful as benchmark guides to past performance which is useful for managers to guide their training sessions, transfers or game plans. Attaching any more importance to them, is the folly of a fool.
