# Redefining face-off success using shot data

Face-offs are one of the most frequent and most visible components of a hockey game. But thus far our ability to statistically measure their impact on games, both for individual players and as a whole, has been limited mainly to the face-off percentage (face-off wins divided by face-offs).

The simplicity of this statistic has the unfortunate side effect of treating all face-off wins and all face-offs as equivalent, along with being totally divorced from shooting and scoring events that actually win and lose games.

Each face-off has a binary outcome in NHL score sheets; they get one winner and one loser. An uncontrolled whack over to the boards that your winger has to spend the next 10 seconds battling for – that gets the same win as a pinpoint pass to your defenseman who proceeds to launch a slap shot through traffic and over the goalie’s glove.

If some face-off takers have a knack for generating more pinpoint passes and less puck battles, that is an advantage they provide their teams that goes unnoticed by only looking at the face-off percentage. Using game event data from after the face-off, we can redefine face-off success and produce a new measurement that is more directly tied to goal scoring and goal prevention.

## Not all face-offs are equal

To see the impact of face-offs on the flow of shots, I looked at every shot in 2013-14 (both shots on goal and missed shots, i.e. “Fenwick” shots) that came after play began with an even-strength face-off. I also tracked how many seconds had passed since the last face-off for each shot, which zone that face-off was in, and whether the shooting team had won the face-off. Using this data I calculated the net shot flow after face-offs (shots by the face-off winning team minus shots by the face-off losing team) for the league as a whole.

Judging by the very different shapes of the lines above, face-off wins in different zones generate very different outcomes. A win in the offensive zone generates many shots and a win in the defensive zone is very significant for preventing shots. And while a win in the neutral zone does produce an advantage, it’s comparatively miniscule.

That flurry of shot activity after zone face-offs appears to last about 10 seconds. (If the offensive zone team won, they continue to have a shot flow advantage through 20-25 seconds, but the vast majority of that advantage has diminished by the 10 second mark.) And if the defensive zone team won, shot flow is essentially even with respect to the face-off at 10 seconds; the defensive face-off win has neutralized the defensive zone start.

## Capturing face-off win skill

So let’s go with 10 seconds as the window of heavy face-off influence on gameplay. Now we can attempt to capture face-off winning skill for a player by looking at all shots that occurred within 10 seconds of that player’s even-strength zone face-offs. Since neutral zone face-offs don’t generate much of a shot flow advantage, they won’t tell us much, so we’ll look only at offensive and defensive zone face-offs for now.

After accumulating all shots that occur within 10 seconds of a player’s zone face-offs, I’ll turn the net shot totals into rate statistics by dividing by the player’s number of face-offs in that zone. This not only facilitates comparison between players with varying numbers of face-offs, but respects the inherent advantage of the zone start by crediting players for defensive zone face-offs resulting in no shots-against (i.e. zero shots is a good thing) and penalizing players for offensive zone face-offs producing no shots-for (i.e. no zero shots is a bad thing).

## Net Shots Post Face-off

The result is Net Shots Post Face-off (NSPF), the balance of shots on goal and missed shots (“Fenwick” shots) in the 10 seconds following even-strength zone face-offs, per even-strength zone face-off.

Here are your offensive and defensive zone NSPF leaders for 2014-15 through January 29 (minimum 100 zone face-offs). I’ve included the even-strength zone face-off percentage as well for comparison. (You can view full, up-to-date NSPF offensive zone and defensive zone data tables at my site, faceoffs.net)

## Goals Post Face-off

For all this shooting, it turns out that goals within 10 seconds of a face-off are not extremely common. Since 2009-10, about one in 120 even-strength zone face-offs results in a goal within 10 seconds, and every game has about 29 even-strength zone face-offs. Do the math and goals after even-strength zone face-offs are a once in four games occurrence – not really enough to make or break a season but frequent enough to make the difference in a few close games.

The small number of goals is why we look at shot data in NSPF above, but if we use multiple seasons to get a larger data set, the numbers for Net Goals Post Face-off (NGPF) get a bit more interesting. Here are your NGPF leaders for the time span 2009-10 through 2014-15 as of January 29.

## Considerations

Before running with post face-off event statistics as a face-off performance measurement, there are some things to consider in addition to the aforementioned minor relationship between even-strength zone face-offs and scoring.

First, so far we have not controlled in any way for context. A face-off taker will benefit in this measure by having more talented linemates who can better generate shots. A future version of these statistics may account for this, but for now it’s worth keeping in mind the player’s context while reviewing numbers.

Second, we have assuming that the outcome of the face-off is totally responsible for the events that occur within 10 seconds; you can certainly have a sequence of events like a clean face-off win followed by a giveaway, where the face-off taker would be wrongly credited. But hopefully the impact of these events would be diminished with a large enough sample.

Third, even if all external circumstances were accounted for, if a player just has a better talent for generating shots irrespective of his face-off skills, then this measurement is going to reflect that a little bit. But that might not be such a bad thing; if a player is an excellent face-off specialist but struggles in other aspects of the game, it’s worthwhile to consider that he turns into a major liability once the puck hits the ice. After all, part of being chosen to take a face-off is being dropped right in the middle of the action and that’s a worthwhile basis on which to choose your face-off taker.

## Conclusion

Post face-off event statistics are a first attempt at using event data to measure face-off skill in a new way, extending beyond the simple face-off. While the impact of even strength face-off performance on team scoring is small, it could be the difference for a team on the edge. Any comments or suggestions would be appreciated.

## 9 thoughts on “Redefining face-off success using shot data”

1. Good stuff but I don’t like the fact that you constantly suggest that this kind of work hasn’t been done before. I first started doing this kind of stuff 3 years ago and Tyler Dellow did a lot of this stuff last winter/spring (though somewhat incorrectly IMO). Others have done it too. While this is a slightly different take on the subject it is hardly new.

• Hi David,
I definitely know better than to claim that anything has never been done before, especially in a field like this where so much of the past research went offline before I had a chance to see it. Nothing in this article was meant to suggest that.

The only thing I meant to suggest is that if a fan wanted a statistic to evaluate current season face-off performance the choices were either face-off win derived metrics (which 99% of the time is straight face-off percentage) or spinning up something else like I did here.

I never came across any of your work while searching and I wish I had because I’m sure it would have helped me a lot, so please link me to it if you can so I can use it to steer my future efforts on this topic.