Thoughts on the Lacrosse Reference preview of Notre Dame-Denver

There are a couple of parts of the Lacrosse Reference preview of the Notre Dame-Denver game that I wanted to comment on. There is some good information there and it gets many things right, but there were a few things that I disagree with and think are misleading.

The intent isn’t to bash them or try to say that they do bad work, but criticism is the only way that lacrosse analytics work can be improved and make progress.

First, on looking back to Denver’s loss to UNC, they write:

In that Carolina win, Balkam had 17 saves. Denver had 13 face offs wins, but was only about to convert three of those into goals. The Heels scored 3 goals on 8 face off win possessions. True to form, Denver only had 6 turnovers. They scored their 9 goals on a whopping 43 shots.
It was the kind of game where North Carolina sat back in defense, tried to prevent easy goals, and hoped that their keeper was up to the task.
It’s a model that Notre Dame could use to beat the Pioneers.

If you’re going to look back to UNC’s win over Denver for a blueprint on how to beat the Pios, defense is the one area Notre Dame was better.

Their stats have Denver as scoring on 9 of 29 possessions (31%) against UNC while only scoring on 11 of 41 possessions (26%) against Notre Dame. My own numbers differ slightly, but still show that the Notre Dame defense was more successful than UNC on a per possession basis. The problem was that the Notre Dame let Denver have too many extra possessions and didn’t score on enough of their own. There is always room for improvement, but I have only Ohio State and Marquette (in their win) holding Denver to a lower offensive efficiency.

Secondly, with regards to Notre Dame’s face-off strategy, they wrote:

In other words, do you all but concede the face off in order to minimize the number of goals Denver is able to score off of those possessions. The numbers say that might not be such a bad idea.
The chart shows the Notre Dame defense’s efficiency marks after face off losses for each game. In games where they’ve allowed the offense to score on 26% of those possessions, they have exactly one win. In games where the opponent is under 26%, they are undefeated.

There is a chart that I will make you go look at the actual article to see and then they continued:

Denver has tended to be somewhat of an all or nothing offense, even more so in their first game against the Irish. In that one, they scored 7 goals on 18 face off possessions and just 4 goals across their other 23 possessions. If you are Notre Dame, it’s got to be tempting to say: “screw it, we aren’t going to win these face offs so might as well batten down the hatches and prevent face-offs from turning into goals.”

There might be some insights to be gained from breaking down efficiency into the different ways that possessions begin, but this is a case where it’s incredibly misleading. If you go through and watch the film, the idea that there is something unique about possessions Denver begins with a face-off win that make them especially threatening to the Notre Dame defense quickly falls apart.

There is one goal that Denver scores directly from a face-off win, but it’s a case where Danny Logan emerges with the ball from a 3v3 scrum and moves the ball to Austin French who hitches and curls topside:

Of the other 6, all you need to know about 3 of them is that they were man-up goals and a fourth goal came after a man-up opportunity had expired. Of the penalties, only the one for 3 face-off violations in a half even had anything to do with face-offs.

The other two were sort of related to face-offs in that they were scored while Denver was still in the middle of subbing off their face-off personnel, but in both cases, Notre Dame already had the 6 players they wanted out on the field to play defense:

In both cases, it’s really a stretch to say that the face-off had really anything to do with the goal from a strategic standpoint. Certainly Notre Dame isn’t going to be modifying their face-off strategy to drop wings back on defense faster to prevent those types of goals from being scored.

I think this situation has 2 important lessons:

  1. Analytics can’t function in a vacuum. It’s important to go back and check film to make sure that you think the numbers say is actually reflective of what is happening on the field.
  2. Be careful about drawing conclusions from small sample sizes or reading too much into correlations and connections between things without doing the empirical work to establish the importance first and considering the possibility that there are other, better explanations.

In this case, there is undoubtedly a correlation between the success of Notre Dame’s defense after they lose a face-off and their ability to win games. But, they only won 46% of their face-offs and so a decent number of their defensive possessions start with a face-off loss. There is a fairly obvious connection between how well a team’s defense plays and whether or not they win a game.

I can replicate the exact same correlation between defensive efficiency and scoring margin using all defensive possessions, not just ones that start with a face-off loss:

The magic number for Notre Dame’s defense in all but 1 game has been about 31% as they’ve won every game where they’ve held teams below that efficiency and only won once when the other team was above it (beating UNC 14–13 in the regular season). Other than their 5–4 win over Maryland where it was 11.8%, between 25% and 50% of Notre Dame’s opponents’ offensive possessions have started via face-off wins. Thus it shouldn’t really be a surprise that defensive efficiency on those possessions shows a correlation to winning similar to defensive efficiency on all possessions.

Further, if most of the possessions that start with a face-off win end up looking like any other possession, then the efficiency on those possessions will be correlated to the efficiency on all other possessions making it appear as though defense after face-off losses is predictive of game outcomes when it might really just be the result of defense in general.

A better test would be to use all other possessions are a control for general defense and look at the difference between that efficiency and efficiency after face-off losses. I think the correlation between that and winning games would start to look much less predictive.