Wednesday 4 February 2015

The Myth of the Eternal Goalkeeper.

It is not controversial to say that goalkeepers are slightly atypical of footballers in general. They have come in all shapes, but mostly one size (tall), appear to disproportionately prefer kicking with their left foot (although this is a personal observation that may not be backed by data), but most strikingly they appear to maintain their abilities into middle age.

Of the 19 seasons played by a footballer aged 40 or greater at mid season since 2002, 14 have been recorded to keepers. Many of these keepers have been solid regulars in their forties, rather than occasional beneficiaries of squad rotation or injury to a first choice alternative. 40 year olds accounted for 0.8% of the total seasons played by all keepers from 2002, but amassed 1.1% of the playing time for all keepers.

So the likes of van der Sar, Schwarzer and Friedel punched well above their weight/age and this has led to the common belief that goalkeeping is an eternal talent.

Of course, it is not true. Even the likes of Friedel have eventually called time on their top flight career. But their obvious presence in the lineup helps to perpetuate the myth that all keepers are age resistant, whereas attrition rates paint a very different picture. Of all keepers who played in the Premier League since 2002 around 90% of them had hung up their boots before the age of 40.

Friedel at 40 may have been good enough to take playing time from a younger, less talented alternative, but if we want to measure age related decline, especially in older groups of players, we need to compare Friedel to his younger self, rather than to a younger alternative player.

We did this in this previous post. Charting the change in playing time with age for individual players and then combining these changes for an overall age related decline or rise in allotted time.

Unlike baseball, where games and individual repetitive trials for the players are available in large numbers, an objective key performance indicator is largely unavailable for football. Therefore, change in playing time is used as a subjective, but informed proxy for talent.

The curve plotting increase or decrease in playing time charts the gradual shedding of appearance time by each keeper in the sample once they reach a subjective peak, including those 90% who fell by the wayside earlier than the likes of Friedel and Schwarzer.

Keepers appear to peak later than other positions. The previous post suggests 29 as a peak age. And so for completeness I've included similar plots for defenders midfielders and strikers.

At the moment players have been broadly categorised as defenders, midfielders and strikers, although further subdivision within these groups is obvious and desirable and the straight line plot would probably be better suited by a curved one. Similarly, a single, regressed figure to denote player talent or a single aspect of a player's ability would be preferable to the use of playing time.

But this approach does attempt to address the problem presented by the steady accumulation of more talented, but declining players populating the older aged groups in all positions and perhaps creating a more optimistic projection for older talent than is the case.

The age at which players in general from each positional group begin to lose playing time is 25 for strikers, 26 for defenders and 26.6 for midfielders.

Data on playing time has kindly been provided by Infostrada Sports with the help of Simon Gleave.

1 comment:

  1. Interesting analysis! How do you handle debut seasons (and debut age)? While many young players progress as Sterling has, with a 'cameo' season followed by a starting role, transfers from foreign leagues may go from 0 to thousands of minutes, which would give us a dividing-by-0 percentage change.
    As for debut age, while studies on baseball have explored the connection between age-at-debut and talent level, I'm not familiar with similar work for football -- if a player first plays in the PL at age 20, do they receive 0s for their age-17 through age-19 seasons for these analyses? This is just intuition, but the playing-time increases pre-peak seem smaller than I would have expected.