Tim Sherwood, in the stale boilerplate of club speak, “has left his coaching position at the Club” according to the Tottenham Twitter account (apologies if that’s you Gary Andrews).
Shortly after the announcement, the popular OptaJoe account mentioned Tim’s win percentage as manager:
59% - Tim Sherwood had the highest winning percentage of any Tottenham Hotspur manager in Premier League history. Gilet.
— OptaJoe (@OptaJoe) May 13, 2014
Which was the subject of some consternation from analytics writer Ted Knutson, who rightly pointed out that a Point Per Game metric is vastly superior to win %.
If it’s not obvious, the reason involves the 3 points for a win, 1 point for a draw system, which win % tosses out (incidentally Sherwood is top of the list there, too). The other issue (which affects both metrics in fact) is the small sample of games involved. Win percentages almost always dip the longer a manager is at a particular club. If you’d like a ridiculous example, Vincente Del Bosque has technically the highest win percentage in Real Madrid’s history during any single managerial stint (he managed more than once), but that involved only one game in 1996.
Okay, so why does any of this matter?
It’s easy to rag on fans, journalists, whomever, when citing bad stats. Once they’ve ingrained in public consciousness, they’re very hard to dislodge. Despite years of Sabermetric progress in baseball, most TV networks and online boxscores still feature RBIs and pitcher win/loss records.
Baseball, like soccer, has several strong voices who’ve spent years, sometimes decades, making detailed arguments for more effective statistics in their sport.
The difference is that baseball has searchable databases like FanGraphs for fans interested in making their own conclusions over various teams and players based on the latest advanced statistics.
Football, on the other hand, has a handful of sites that presents raw data collected by Opta. There you will find nothing on Non Penalty Goals Per 90, nothing on Total Shots Ratio and PDO, nothing on Expected Goals.
Yes, there are various sites which feature these numbers, but they are usually static tables manually uploaded by an often busy author. They’re also scattered across a number of different websites, some of which are more easily searched out than others.
Whenever I’ve raised this with other soccer analytics people, the issue of proprietary stats comes up. Yet some of the most power statistics don’t require in-depth data to calculate, with the exception of Expected Goals. Most of it is based on shots and goals. Hell, I may try and do it myself.
What does this have to do with Tim Sherwood and win percentage? After reading and writing on football analytics and statistics for the last couple of years, I’ve discovered the main issue isn’t the use of stats per se, but the prevalence and availability of bad statistics.
Check any Wikipedia page on any manager for example and there is almost invariably a win percentage table.
Nothing on points per game.
This could easily be rectified with some nerdy editing.
The broader point however is that changing how the public perceives and uses sports statistics is a long process. The first, most important step is making this data available in once place, automatically updating it, and making it easily searchable.
Soccer isn’t there yet. Until it is, railing against the general public for not understanding the difference between win% and PPG is less effective than simply making better stats more easily available.
