Stats updates – 5vs5 and more

It’s been a while since my last blog post but I’ve been quietly working behind the scenes to set the stage for some new features to come. I’ll be rolling it out bit by bit.

For now, just one little thing: I am adopting the “industry standard” of using 5-vs-5 data when presenting my NSPF advanced face-off stats, rather than using all even strength situations as before (which would include 4-vs-4 and 3-vs-3).

So the numbers may look a little different than a couple days ago as I am not counting 4-vs-4 and 3-vs-3 anymore, but hopefully this gives a better basis for comparison.

I am always looking for suggestions on how to make this site more useful; please reach out if there is something specific you are looking for.

NSPF: Improvements to the shot-based face-off metric

Patrice Bergeron wins another face-off

The best way to estimate face-off ability in the NHL, in my opinion, is with my statistic NSPF (Net Shots Post-Faceoff). NSPF estimates a player’s ability to win face-offs, and more importantly, his ability to influence shot flow, by looking at how often shots occur during the 10 seconds following his face-offs. You can read more about why I use 10 seconds in my original introduction of NSPF at Hockey Prospectus.

Since the introduction I’ve made some small changes to NSPF. Specifically:

  • I’m now basing the numbers by default on all shot attempts (previously it was unblocked shots only), although on the stat page you can switch between other shot types like unblocked shots, shots on goal, and goals.
  • I’m now including face-offs in the neutral zone.
  • I’m now adjusting the numbers in comparison to the league average. Special thanks to Rob Vollman for suggesting this.

Here’s how NSPF is now calculated from start to finish, using Patrice Bergeron‘s 2014-15 regular season as an example:

First, we count the number of shots (for and against) during the 10 seconds following the puck drop on all of Bergeron’s even-strength face-offs, according to whether the face-off was in the offensive zone, defensive zone, or neutral zone:

(A) (B) (C) (D)
Zone Bergeron
Shots For
Shots Against
Net Shots
(B) minus (C)
Offensive 412 136 4 132
Defensive 534 14 87 -73
Neutral 478 20 11 9

While it’s nice to know his net shots, it doesn’t tell us by itself how much better he is than other players in the league. Almost all players will have positive net shots in the offensive zone, negative net shots in the defensive zone, and close to zero in the neutral zone.

So now we form a basis of comparison to the rest of the league by calculating the amount of net shots we would expect Bergeron to have, which is the league average rate of net shots after face-offs times the number of face-offs he took.

For the neutral zone that’s easy; the net shot rate is zero across the league by definition, because every shot-for counts as a shot-against for someone else’s neutral zone face-off. For the other two zones we’ll find the shots-after-faceoffs rate by calculating the total number of shots in the 10 seconds after face-offs (minus Bergeron’s) and the total number of zone face-offs (minus Bergeron’s).

Note that the league values for offensive and defensive zone net shots are equal in magnitude and opposite in sign; like with the neutral zone, everyone’s shot-for is someone else’s shot-against.

(E) (F) (G) (H)
Zone League Face-offs League Face-offs
minus Bergeron
League Net Shots League Net Shots
minus Bergeron
(E) minus (A) (G) minus (D)
Offensive 39,391 38,979 8,985 8,849
Defensive 39,391 38,857 -8,985 -8,912

We use these numbers to form a shot-per-face-off ratio that we multiply by the number of Bergeron’s zone face-offs to obtain the expected net shots. With the expected value we can come up with an adjusted number that reflects Bergeron’s ability in relation to the rest of the league.

(I) (J) (K)
Zone League Shots per Face-off
(minus Bergeron)
Expected Net Shots
Adjusted Net Shots
(H) div. by (F) (I) Ă— (A) (D) minus (J)
Offensive 0.2270 93.6 38.4
Defensive -0.2294 -122.5 49.5
Neutral 0 0 9

With the adjusted net shots calculated per zone, the total adjusted Net Shots Post-Faceoff is simply the sum of the numbers in column K: 96.9

To form a basis of comparison with other players with other quantities of face-offs, we can divide that number by Bergeron’s even-strength face-off count, 1424, to get his NSPF per face-off: 0.0680.

While it might be hard to ascribe meaning to a small number like that, here’s a simple way to do it. 0.0680 is approximately 1/15. So every 15 face-offs at even-strength, Bergeron’s face-off taking ability compared to an average player gains his team an additional shot attempt (or denies the opposing team a shot attempt) within 10 seconds of the face-off. For a player who averaged 17.6 even-strength face-offs per game last year, that’s a nice bonus for the Boston Bruins. And that’s just compared to an average player; his value compared to a below-average player would obviously be even greater.

However I do want to emphasize that NSPF is an just estimate of a player’s face-off winning ability, because we are judging a player by all shots that occur within 10 seconds (whether or not the face-off had any bearing on them) and ignoring any events that occur after 10 seconds. We do so because statistically the window of face-offs influencing shot flow has been shown to be strongest within 10 seconds, and that’s the best we can do at the moment. When advanced player tracking lets us apply a better filter, we’ll come up with a better way to judge face-off talent by the events that follow.

Photo by Erica Britto, used under CC BY 2.0

Proposed defensive zone face-off rule would have miniscule effect on scoring

In its ongoing quest to increase goal scoring, the NHL/NHLPA Competition Committee has proposed a rule change to modify defensive zone face-offs so that the defensive zone player must always put his stick down first. This replaces the current format in which the visiting team player puts his stick down first. (Neutral zone face-offs would continue to follow the existing rule.)

It’s worth noting that over the past five seasons, the offensive zone team has only won face-offs at even-strength 48.4% of the time. Maybe the league has noticed and wants to bump that number up to 50% or higher, and thinks that a rise in scoring would follow.

But would this rule change actually cause a significant increase in scoring? I decided to do the math and the result was only a miniscule increase in goals. Here is my methodology:

  • I looked at every goal scored in the NHL from 2010-11 through 2014-15 and the outcome of the last face-off prior to that goal.
  • I calculated the league-wide rate at which goals were scored (i.e. goals per face-off) by either team after face-off wins and losses in each of the following situations: (1) even-strength offensive zone, (2) power-play offensive zone/shorthanded defensive zone, (3) shorthanded offensive zone/power-play defensive zone.
  • During these five seasons the home team won even-strength face-offs 51.6% of the time (or, a 1.6% advantage compared to a coin flip). While there are surely several factors contributing to this home-ice advantage, for the purpose of finding the maximum effect of the proposed rule change, I attributed the full 1.6% to the current rule forcing the visiting player to put his stick down first, and used this as an analog for the proposed rule.
  • I set up a model of scoring per game based on the number of face-offs per game in each of the above situations and the observed rate of scoring after each face-off, according to whether the offensive zone team won the face-off. I adjusted the offensive zone face-off win percentage to be 1.6% higher in each situation as a result of implementing the proposed rule change, and calculated the resulting change in scoring.
  • Here is all of this in a spreadsheet if you want to check my numbers or try plugging in some other number besides 1.6%. (Note that it won’t change the result much.)

So here is the change in scoring per game:

FO/game Change in goals per face-off after adjustment Change in goals per game
Even-strength Off. zone 29.84 0.000113 0.00337
Power-play Off. zone 2.384 0.000307 0.000731
Shorthanded Off. zone 0.639 -0.000233 -0.000149
Sum 0.00395

Multiply that goals per game sum by the 1,230 games in a full season and we are talking about approximately 4.9 additional goals per year across the entire league. It’s an almost unnoticeable difference, but in a league desperate to increase goal scoring, it may be better than nothing.