Analyzing NLL Shooting Numbers

While I was working on the article that I wrote about the value of secondary assists in the NLL, I happened to notice something intriguing about the year-to-year correlations for goals relative to shots that I think is worth sharing with the lacrosse analytics community.

Compare the R² values (coefficients of determination) for goals and total shots:

Note: All of the graphs that appear below use the same data set and methodology as the ones I described in that article.

As you can see, the correlation for total shots per game is substantially stronger than the one for goals per game.

Further there is still a significantly stronger correlation for shots relative to goals even when total shots are broken down into shots on goal and shots off goal.

In fact, the R² value for shots on goal (0.568) is nearly as large as the one for total shots (0.598). Of course, that is to be somewhat expected because of the fact that shots on goal constitute the majority of total shots.

As you may have realized by now, since every goal is the result of a shot on goal and the correlation for total shots and shots on goal is substantially stronger than for goals, that implies that there is substantial noise or randomness introduced by the question of whether or not the goalie makes the save.

And in fact if we look at the correlation for shooting percentages and shooting percentage for shots on goal, they’re essentially non-existent. (“Shooting percentage for shots on goal” is calculated by simply dividing goals by SOG. I sometimes refer to it “1 minus the opposing goalie’s save percentage” because that’s another way of calculating it and phrasing it that way illustrates it’s connection to opposing goalie’s save percentage against that particular shooter. It seems logical to expressed it this way rather than just as the opposing goalie’s save percentage so that higher is better and it’s more like shooting percentage. There should definitely be a better name for this stat and I don’t know why there isn’t one.)

In aggregate, what these numbers suggest is that shots and/or shots on goal are more steady indicators of individual offensive success than goals scored. For whatever reason, it appears as though players by virtue of their talent offensive skill, role on the team, decisions make by teammates and coaches, etc. have a great deal of influence over how many shots they put on goal, but whether or not those shots end up going in might depend mostly on some combination of the opposing goalie and luck or randomness.

For followers of hockey analytics those last couple of sentences should probably remind you of Corsi and Fenwick. I don’t yet have a specific proposal, but it seems like a box lacrosse variation of those numbers or perhaps PDO would be useful tools. (Also, since all 3 of those numbers are named after the person who came up with them, if I come up with something, I feel as though it’s only appropriate to call it “McEwen”.)

Fortunately, unlike second assists, these ideas can be evaluated from the perspective of goalies as well as on the team level and more work can be done see if there is any difference between shots on goal as compared to shots off goal or total shots. Thus, I think we can get a much better grasp of what’s actually going on before making any proposals for new metrics.