Stats

Stuff Maybe Only I Care About

Some interesting data (I think) from recent seasons in the Centennial Conference men's soccer competition:

Each dot is labelled with the school initials or abbreviation and a hyphen year (HC-11 = Haverford College, 2011); hover mouse to see each data point label.\
UC: Ursinus College
DC: Dickinson
MC: Muhlenberg
F&M: Franklin and Marshall
SWAT: Swarthmore
MCD: McDaniel
WAC: Washington College
HC: Haverford
GC: Gettysburg
JHU: Johns Hopkins University

What is "TSR?"

The formula is simple:
-Shots For- divided by -(Shots For + Shots Against)-

The ratio results in an indicator of how effective a team is at getting shots at the one end, and preventing them at the other.  A simple concept.  Among a few metrics, this particular ratio shows as high a degree of efficacy even after a relatively small sample size in predicting performance.  Somewhere around 85% accuracy.  Which isn't great...but that's as good as analytics gets these days in football.  Clearly, there's a trend here, as you might expect; if I give up fewer shots than I take, I ought to win more than I lose.  Seems intuitive.

The main trick is that every shot is just a loss of possession...how does a team take more shots than their opponent without conceding as many?  So the metric becomes a useful indicator of a team's ability to take the ball away prior to a shot by the opponent.  The slop in the data (that pesky 15% or so of unaccounted-for error) comes from teams who are simply off the charts in goal scoring efficiency.  Broadly speaking, one in every 10 shots is a goal (that's shots of all kinds, regardless of on-target, blocked, saved, etc.) from free play.  A team that exceeds the 10% rule can have a wacky TSR for their win percentage.  Man U in 2012-13 (Fergie's last season in charge) or Simeone's Athletico in 2013-14 come to mind.  Of course, there's some similar effect from the bottom of the tables, where teams struggle to finish, though at the pro level the problem isn't in the scoring, it's in getting the ball to the attacking end, or, indeed, getting the ball at all.

At the top level, too, we tend to see teams that spend a lot of time in the lead content to absorb pressure and attack a little less (to reduce the chance of being caught with too many players forward) and so teams that score early (and win a lot) may have surprisingly low shot totals...but if you finish chances at a high rate, the other team has to come after you with more vigor to get back in the game.  Now the second and third goals can come on the counter - a very high percentage situation.

At the amateur level (college) a team that takes a lead tends to continue to attack.  By virtue of being the better team, they press on because they can.  The NCAA Division I playoffs tend to look more professional in this regard, but DII and DIII tend to have more lopsided contests, with a clearly better team present in most games.  Amherst, this year in DIII was a lower-league outlier, with an 8% shooting percentage, made the sweet 16.  They play in a highly competitive league but are one of the stronger teams.  They defend wonderfully and often play with a lead- so they take fewer shots and finish fewer as well.  Heaven help the rest of the NESCAC if they get a forward who's more effective in counterattack conditions than the players they have now.  And their All-American GK has to graduate someday...

At the college level, the ratio of 1-in-10 pretty well holds...but the bell curve will be more dramatic; more teams, a much wider variety of talent.  The median for DI/DII/DIII in terms of shooting percentage is 10.5/11.75/11.5 percent.

Here's the proof from 2014 Men's NCAA soccer: