# How important is neutral zone play?

Eric T.
July 09 2012 11:48AM

A couple of weeks ago, I laid out an overview of Broad Street Hockey's zone entry project. I described some of the cool things we've found and called for readers to join the project, but I didn't yet show any of the data to support my claims.

This article will be step one in that journey; we'll delve into just how much of a team's results are determined in the neutral zone.

As a reminder, zone entry tracking means we note each time a team sends the puck into the offensive zone (with the intent of generating offense; we exclude plays where they just dump it in and go for a line change). We note which player sent the puck in and how they did it -- by dumping it, carrying it, etc. The result is that we can calculate how many shots, scoring chances, and goals are generated from each entry. Here's what we find:

 Carry-in Pass-in Dump-in Deflect-in Shots per entry 0.57 0.56 0.25 0.26 Goals per entry 0.039 0.035 0.014 0.017

It appears that how a team gets the puck into the zone is as important as how often they do it. Maintaining possession of the puck at the blue line (carrying or passing the puck across the line) means a team will generate more than twice as much offense as playing dump and chase.

The common defense of dumping the puck in is that it is the sound defensive play, forcing the opponent to go the length of the ice. But while I need to update things with the full season of data, through 1/4 of the season there was no evidence that carrying the puck in was riskier than dumping it, so we will neglect that possibility for the purposes of this analysis.

We can then tease apart how a team's play in each zone contributed to their shot differential -- which, of course, ties closely to puck possession, zone time, scoring chances, and goals.

This graphic shows how we can roll together various statistics to calculate a performance score for each zone. For example, if a team gets an above-average number of shots per entry, they will have a positive offensive zone score. We can perform these calculations with a given player on the ice to see whether the team performed better in a given zone when he was on the ice.

 Forward OZ score NZ score DZ score Jagr -0.9% +2.6% -0.7% Giroux +0.2% +2.2% -0.7% Read -1.1% +2.0% +1.2% Hartnell +0.4% +1.3% -1.6% Couturier +0.5% +1.1% +2.3% Talbot -0.4% +0.4% +1.1% Voracek +1.2% +0.2% +0.1% van Riemsdyk +1.2% -0.9% +1.5% Simmonds -0.8% -1.5% -0.5% Schenn +0.8% -2.1% -1.0% Rinaldo -0.9% -2.4% +0.9% Briere -0.8% -2.7% -0.4%
 Defenseman OZ score NZ score DZ score Timonen +1.5% +0.9% +1.5% Carle -0.8% +0.8% +0.5% Bourdon -0.7% +0.5% +0.9% Meszaros -0.7% +0.0% -0.9% Gustafsson -1.0% -0.2% +0.6% Coburn +0.0% -0.5% +0.5% Lilja +0.0% -0.6% -0.9% Grossmann +0.1% -0.9% -2.1%

Some things here make sense. It's not hard to believe that Sean Couturier was the Flyers' best forward in the defensive zone, or that Kimmo Timonen was their best defenseman in all three zones.

But at least as many of these results stand out as being surprising. Would you expect Daniel Briere and Jaromir Jagr to be just as inefficient in the offensive zone as Zac Rinaldo? Would you expect Nicklas Grossmann to be the second-best defenseman in the offensive zone but by far the worst in the defensive zone?

When numbers are wildly out of line with expectations, that can mean there's an exciting discovery coming, that our intuition is wrong. It can also mean that the numbers just aren't particularly meaningful, that they aren't really measuring what we thought.

One of the most common ways to take a first glance at whether a metric is meaningful is called a split-half reliability test. We look at how well one half of the data (the odd-numbered games, for example) predicts what happened in the other half. If you can't look at the results in the odd-numbered games and guess what happened in the even-numbered games, then whatever you're looking at probably isn't really measuring a true talent; it's mostly random numbers generated by statistical noise.

The net result here: our surprising results in the offensive and defensive zones appear to be based on a not-particularly-reproducible metric. It may still be true that some players have skills that help the team get more shots per zone entry, but at the end of a year we can't reliably tell which players those are -- we know who did well in the offensive zone this year, but don't have strong reason to believe they'll do well in the offensive zone again next year.

The neutral zone is a different story. The split-half reliability there is 0.44, high enough that we can be 97% sure that this is a real correlation and not just random results. Given a half-season of neutral zone data, we can make a decent guess at what will happen in the other half-season.

Moreover, the neutral zone results look like they are not just statistically significant, but meaningful in practice as well. It's pretty reasonable to see Jagr and Claude Giroux leading the forwards and Timonen and Matt Carle leading the defensemen.

## Moving Forward...

So I'm ready to believe that neutral zone performance is a repeatable measure of a real talent that the players have, and that the offensive and defensive zone scores we've calculated might not be. But we still haven't really answered the question posed in the title of this article: How important is neutral zone play?

It is still possible that neutral zone score (like hits or fighting majors) is reproducible but doesn't have much impact on the outcome of the game. And it's possible that offensive and defensive zone score (like shooting percentage or save percentage) has a huge impact on the outcome but fluctuates quite a bit.

We know that outshooting the opponent is important. And we know it is a reproducible skill, so it seems unlikely that it would be driven heavily by the irreproducible attack zone results. But we can confirm this in practice by looking at how strongly our components relate to the overall shot differential.

The neutral zone score alone explains twice as much of the spread in shot differential as the offensive zone score does, and ten times as much as the defensive zone score does. These factors alone do not completely determine a player's shot differential; obviously how often a player is used for an offensive zone faceoff is a factor, as is the team's performance immediately following the faceoff (compare Couturier getting 0.45 shots between the faceoff and a clear with Wayne Simmonds getting 0.31 shots).

Still, at least for this one season, at least for this one team, we can predict a player's overall shot differential with reasonable accuracy (+/- 0.7%) without knowing how he did in the offensive zone, how he did in the defensive zone, where he got deployed, or how he did on faceoffs. This isn't just because someone who's good in the neutral zone is probably good at those other things too -- the cross-correlations between these metrics were all below 0.15.

At least for this one team, neutral zone performance is the major driver of overall shot differential, which in turn drives results. What remains to be shown is whether this is a quirk of a small dataset or something that will prove to be true as we collect more zone entry data, but for now we will tentatively suggest that the neutral zone may be a much bigger factor in a team's performance than had previously been supposed.

Eric T. writes for NHL Numbers and Broad Street Hockey. His work generally focuses on analytical investigations and covers all phases of the game. You can find him on Twitter as @BSH_EricT.
#1 Ben Bishin
July 10 2012, 12:23AM

Fascinating and very thoughtful. Well done. One thought: the comparison of effectiveness on dump ins versus carry or pass ins might be driven by the possibility that dump ins are more likely to happen on out-maned attacks (ie the defensive team has a man advantage against the rush as in a 2 on 3 attack) while carry and pass ins are more likely to occur on attacks where the offensive team is either even manned or has an advantage going in in to the offensive zone (e.g., a 3 on 2 attack). Of course these are situations where we would expect the offensive team to score more often due to the man advantage. So it is possible that the choice of how to enter the zone is a function of strategic situation the puck carrier faces.

Its not clear to me that this effects your broader findings in terms of the overall zone analysis, but I suspect it will lower the gaps between the relative scoring effectiveness of the different ways of carrying the puck in to the zone.

#3 Clayton
July 10 2012, 08:22AM

Did you do a comparison with score? Or consider only using score-neutral stats?

I am just assuming you would see a higher number of dump-ins with a lead. The converse would also be true that a team down a goal would carry the puck more and look for more shots.

#5 Kent Wilson
July 09 2012, 11:56AM

Awesome stuff.

If the project continues and the findings are confirmed, this has major repercussions for the way clubs build on-ice strategy and view individual players.

#6 Josh L.
July 09 2012, 12:08PM

Dear god I see Toby Petersen

"Would you expect Nicklas Grossmann to be the second-best defenseman in the offensive zone but by far the worst in the defensive zone?"

Yes. For whatever reason he seems to be terrible in transition.

#7 Derek Jedamski
July 09 2012, 01:44PM

Awesome stuff

#8 draglikepull
July 09 2012, 05:11PM

Back in the winter I was working on a series of posts over at Pension Plan Puppets where I measured the amount of time the Leafs and their opponents spent with control of the puck in the offensive zone (ie. puck possession) (you can find them by searching for "Measuring Puck Possession" at PPP). I was only tracking possession, not goals or shots, but my observations line up very well with what you've found here. I didn't formally track it, but my observation was that teams spent much more time in the attacking zone (and consequently got more shots) if they skated into the zone with possession, while dumping the puck in rarely resulted in good opportunities.

I also found (and I'll go into more detail on this in a couple days when I finish writing my wrap-up post on this topic) that the vast majority of the game is spent either in the neutral zone or in transition. An individual team very rarely spends more than 3-4 minutes of even strength time in a period in the attacking zone (at least in the Leafs games I tracked). So your results, while they may sound a bit unintuitive, do line up with the research I've been doing as well.

#9 Clayton
July 10 2012, 08:22AM

Did you do a comparison with score? Or consider only using score-neutral stats?

I am just assuming you would see a higher number of dump-ins with a lead. The converse would also be true that a team down a goal would carry the puck more and look for more shots.

#11 Ben Wendorf
July 10 2012, 10:11AM

One question: did you tease out the possible effect of dump-and-changes? I could see that bringing the shot/goal results of carry/pass-ins versus dump-ins a bit closer.

#12 Ben Wendorf
July 10 2012, 11:01AM
Eric T. wrote:

Yeah, dump-and-change plays are excluded from the analysis altogether. We only logged offensive plays, where the team with the puck makes an active effort to recover it after dumping.

D'oh, just saw it in there. Hooked on Phonics.

#13 mac
July 10 2012, 03:04PM

Very interesting.

Watching the Sens last year, a team I expected to be terrible, one thing I noticed as a big contrast to the Leafs (who I usually watch) was that they were REALLY good at gaining the zone. One team made the playoffs.. one did not.. :D

One technique the Sens seemed to use a lot was to gain the zone heading into traffic, then just ditch the puck laterally towards the boards for another player gaining the zone to pick up on his way into the zone.

#14 garik16
July 10 2012, 05:50PM

I'm probably missing a linked post up there to your methodology somehow, but how did you separate out the scores between neutral zone, O Zone and D Zone? I see the chart showing what you consider each type of play but not where you show how you measure positive performance in each zone.

#15 garik16
July 10 2012, 05:52PM
garik16 wrote:

I'm probably missing a linked post up there to your methodology somehow, but how did you separate out the scores between neutral zone, O Zone and D Zone? I see the chart showing what you consider each type of play but not where you show how you measure positive performance in each zone.

To build (nhlnumbers really needs an edit comment feature sigh) on this, I see where you say O and D Zone #s are based upon beating the average zone performances per type of entry, but are the neutral zone numbers just you essentially counting the zone entries as resulting in the average # of shots?

#16 garik16
July 10 2012, 06:08PM

I also wish we had a larger sample size on this, because I'd love to see how this plays out by position - the neutral zone thing makes a ton of sense for forwards, but you don't think about D-Men in the neutral zone.

#18 garik16
July 10 2012, 08:43PM

Thanks Eric, I'm pretty sure I get it now, reminds me of how expected run values are used in baseball - you're taking the average results of each type of entry and using that to measure how well someone plays in the neutral zone. (Expected Run Values in baseball measure pitch values by taking the average value of a ball hit in play based upon what batted ball type it is - ground ball, fly ball, or line drive - rather than the actual result).

#19 Kent Wilson
July 11 2012, 10:15AM

@garik16

There is an edit comment feature if you're logged in.

#20 ken
July 11 2012, 06:23PM

"The result is that we can calculate how many shots, scoring chances, and goals are generated from each entry."

The chart following that only has shots and goals. I'm imagining the scoring chances show a similar trend, but I'd guess that it might be a smaller gap than shots (since many carry-ins are glorified dump-ins ending in a weak shot from the boards).