How We’d Fix It: Shooting Percentage

Sports have moved into an era dominated by advanced metrics. Medieval scouting tactics have been replaced. General managers in every sport seek the most accurate metrics for determining player efficiency. Hockey geeks now regard Fenwick and Corsi numbers as predictors of team success. Both numbers have the same concept: Focus on possession rather than scoring, because it provides a larger sample size and eliminates uncertainty.

In baseball, on-base percentage is often accepted as ‘more accurate’ than batting average. Slugging percentage has been used independent of OBP [and with it to form OPS] to measure extra-base hits. Kirk Goldsberry’s ShotScore and John Hollinger’s PER are my two favorite advanced metrics in basketball, but there are plenty more – win shares, true shooting percentage and usage rate.

What do these statistics all have in common? They use instrumental variables that, despite being less correlated with current success, are better predictors of future success. Corsi numbers observe a player’s plus/minus of shots on net rather than goals. Goals are relatively rare in hockey and therefore, can skew data. Although goals literally explain current success, they are poorer indicators of future success than shots.

Shots are better measurements of possession due to their larger sample size, and possession is a solid predictor of team success. Here are the Fenwick Close percentages for all NHL playoff teams since 2009. Teams below the line have percentages better than 50%; teams closer to the dot fared better in the playoffs.

Infographic_playoffNow look at the Fenwick percentages for non-playoff teams. Most are above the 50% line.


On-base percentage focuses on the same concepts. In general the team that puts the most players on base will score the most runs. Batting average, like goals, is the sexier metric; OBP, like shots-on-net, is the more robust statistic. At the end of the day, do general managers care how players get on base?


Currently, lacrosse uses two statistics to measure shooting success – shooting percentage and shots-on-goal percentage (SOG%). Shooting percentage is simply goals divided by shots. SOG% is shots-on-goal divided by shots. Both are useful, yet incomplete. They leave me wanting more.

What if a player has a terrible shooting percentage, but all his missed shots are off-net and backed up? What if a player has a tremendous SOG%, but every shot is cleanly saved by the goalie?

Of course, no single player is going to fit that mold – but every player is going to have missed shots backed up. Here’s my idea: Calculate the shooting percentage of the shots-on-goal. Call it Adjusted Shooting Percentage. It removes shots that result in an offensive rebound from the divisor. Possession shots won’t hurt a players shooting percentage because, well, they shouldn’t.

The closest we can come to this without tracking offensive rebounds would be dividing goals by shots-on-goal. This assumes all shots off-target are backed up and harmless to the offense, which obviously isn’t true. However, most shots are backed up by the offense – so it isn’t completely wrong.

Let’s take a look at some high-volume shooters. Peet Poillon took 83 shots last year. Of those 83, 45 were on goal and 23 were goals. So, is Poillon a 28% shooter? Or is he a 51% shooter?

You might not think it matters. As long as we stick by the same metric for everyone, there’s no difference. Right? Think again. Paul Rabil took 142 (!) shots last year. Of those 142, 99 were on goal and 32 were goals. That makes Rabil a 23% shooter with a 32% adjusted shooting percentage. Who is the better shooter – Poillon or Rabil? By the first measure, it’s close to a tossup. With adjusted shooting percentage, Poillon was a much better shooter this season. [I say this season because Rabil hasn’t always shot such a low adjusted shooting percentage. More on this some other time, though.]

The most impressive group – and probably the best argument for utilizing this statistic – is the Denver Outlaws midfield. This unit is largely responsible for Denver’s 14-0 season. [Much of the credit has to be given to the defense and goaltending, too.] It shouldn’t be surprising that these players benefit the most from adjusting shooting percentage.

Justin Pennington (71%), Justin Turri (69%) and Terry Kimener (69%) posted the top three adjusted shooting percentages among midfielders. Drew Snider (57%), Zack Greer (49%), Jeremy Sieverts (47%) and Will Mangan (44%) were not far behind. Most of Denver’s midfielders were far ahead of the league-median 48% adjusted shooting percentage.

This was no accident. Denver’s use of two-way midfielders helped them generate easy shots. Their incredible adjusted shooting percentages helped them limit the number of saves opposing goalies made and, in turn, limit their opponents’ transition opportunities. Denver extended possessions by missing well. [Yes, some misses are actually good.]

On-base percentage, Fenwick numbers and adjusted shooting percentage are all predictors of offensive success. By no means are they perfect – but they are more useful than batting average, shots-on-goal and raw shooting percentage. While the “primitive” metrics describe what happened, the advanced metrics predict what will happen.

Shooting a high adjusted shooting percentage makes life easy on your defense. Plus, it seems like a great measure of offensive efficiency, right? I’d say so. It should be a decent indicator of a team’s success. But then again, trigger-happy midfielders Michael Kimmel (36%), Steven Brooks (35%), Ben Hunt (32%) and Kyle Dixon (15%) led the Chesapeake Bayhawks to The Jake in 2013. Go figure.

[Note: Krossover, a sports technology company specializing in basketball, football and lacrosse, currently tracks offensive rebounds. Krossover has done exceptional work in the advanced metrics field. Their breakdown of lacrosse is phenomenal. Some of their stats (i.e. clean-save percentage, offensive rebound percentage and shot result distribution) can help create an even more efficient shooting metric. Eventually, I’d like to see MLL compare an individual’s shot location and on-goal location to the league averages to create a super-stat like Goldsberry’s ShotScore. With Krossover, it’s possible.]

Have a question or idea for Joe? E-mail him at! Include your first name and hometown. You could be featured in his next mailbag column!

Comments are closed.