The importance of zone entries in hockey statistical analysis will come as no secret to anyone familiar with the public community at large. Back in 2011, then-Broad Street Hockey writer (and current Carolina Hurricanes manager of analytics) Eric Tulsky initiated a video tracking project that became the first organized foray into the zone entry question, and later resulted in a Sloan Analytics Conference presentation. Tulsky determined that “controlled” entries (those that came with possession of the puck) resulted in more than twice the number of average shots than “uncontrolled” entries, a key finding that provided concrete direction for additional research on the topic.

Tulsky’s initial Sloan project was limited, however, due to lack of data – only two teams had their full regular seasons tracked, and just two others reached the half-season threshold. As a result, further research would wait until a larger dataset became available. Luckily for the community, Corey Sznajder undertook a massive tracking project encompassing the entire 2013-14 season, and released the data to the public. Using this, there were more advances, including Garik16’s work on team zone performance and the repeatability of player performance in each individual zone.

Through that work, the community learned that teams can sustain shot volume outcomes better than expected averages in the offensive and defensive zones, and Tulsky’s original thesis that neutral zone outcomes were repeatable on a team-level was confirmed as well. On the player level, however, Garik determined that on-ice shot volume and prevention results in the attacking zones were not repeatable, and only neutral zone outcomes (weighted entry differential) could be trusted.

On-ice outcomes are just part of the data that can be extracted from zone entry tracking, of course. The easiest metrics to compile when looking at individual players are the ones that measure one’s ability to drive zone entries on their own. A player’s ability to carry or dump the puck into the offensive zone clearly provides value to the team, in the most obvious sense that goals are rarely scored from outside the attacking zone.

But just how valuable are these individual entry generation metrics in evaluating individual player performance? And if they have value, which of the raw metrics should be held in the highest esteem?

Introducing Weighted Entries per 60

Since the data has been tracked, there have been two primary metrics used to evaluate individual players’ ability to drive neutral zone entry creation results. First is Entries per 60, which is a simple measurement of total entries created by a player accounting for ice time in a situation. Second is Controlled Entry Percentage, which is a ratio measuring the number of entries created by a player with possession of the puck versus the total entries created (controlled and uncontrolled).

Both metrics have their inherent weaknesses. Entries/60 does not account for the quality of the entries created; a player with seven dump-in entries in a game will likely have a higher Entries/60 rate than a teammate with five controlled entries, even though the latter’s entries were more likely to produce shots. Controlled Entry Percentage has the opposite problem, as a player with one controlled entry and zero uncontrolled ones would appear at first glance to be more impressive than a rival with a three controlled/one uncontrolled ratio. Clearly, that doesn’t pass the common sense test.

One way to address this issue is to use the framework of the Entries/60 stat, but weigh each entry based on the expected shot-based outcome. Per Sznajder’s tracking in 2013-14, the league average outcome for a controlled entry was an average 0.66 unblocked shot attempts, while uncontrolled entries created about 0.29 unblocked shots. The result is a new metric, which I’ve dubbed “Weighted Entries per 60” or wE/60. The straightforward formula is as follows:

(((Number of Controlled Entries * 0.66) + (Number of Uncontrolled Entries * 0.29))*60) / Total TOI

At its core, wE/60 provides a rough measurement of how many unblocked shot attempts per 60 a player has personally driven via individual neutral zone play, assuming league-average outcomes in the offensive zone. But the real goal is to attempt to combine the strengths of Entries/60 and Controlled Entry Percentage into one metric.

Of course, just because a metric makes theoretical sense doesn’t mean it holds up to testing. For that matter, all three metrics must be tested to see just how valuable they truly are.

Repeatability

Before we move into the usefulness of the metrics, it’s important to confirm that they are actually measuring repeatable skills. Unfortunately, in terms of datasets, we only have a one full season of public manually-tracked entry data, from 2013-14. As a result, a split-half repeatability test of that dataset – checking if the performance of individual players in odd games in a single metric versus their results in even games in that same metric is closely comparable – is the best way to evaluate this. During that season, 689 players skated in at least 200 five-on-five minutes, and they will comprise our dataset.

Additionally, I chose to separate the dataset into two sections – forwards and defensemen. This is because forwards tend to drive substantially more entries than defensemen due to the nature of the position, and therefore checking for repeatability across the full dataset would result in an inflated view of the strength of the relationship. Forwards will, as a group, always generate more and better entries than defensemen considering the prevailing style of play in the NHL, regardless of whether the game is an odd or even number.

With these decisions made, let’s take a look at the R² relationships for each metric and evaluate whether they are strong enough to continue forward with our analysis. For those reading who aren’t familiar with the test, an R² of 1.0 would mean that the players’ results in both the odd and even numbered games were identical, so the closer to that mark that each R² test falls, the more repeatable the metric is.

	Controlled Entry Percentage	Entries/60	wE/60
Forwards	0.7219	0.5566	0.6438
Defensemen	0.5000	0.3959	0.5366

All three metrics show solid repeatability, particularly for the forwards in our dataset. But even for defensemen, entry generation looks like a sustainable skill. Interestingly enough, the raw Entries per 60 statistic performs the worst for both positions, and (spoiler alert!) this just might prove to be something of a trend.

Entry prowess linked with shot generation, especially for forwards

Now that we know that each of our three individual player metrics are repeatable, we can check to see if they are also correlated with positive on-ice outcomes. Since Corey’s dataset utilizes Fenwick (unblocked shot attempts) as its base form of measurement, let’s use it to determine if neutral zone entry generation shows a link to shot creation, shot suppression, or shot differential.

We’ll start with the forwards. Again, we’ll be measuring the strength of the R², but in this case, it won’t be checking for repeatability. Instead, we’ll be checking the relationship between each of our three entry metrics (CE%, E/60 and wE/60) with Fenwick For per 60, Fenwick Against per 60, and Fenwick For Percentage. If the full-season entry metrics show a strong correlation with the full-season shot metrics, we can conclude that a positive relationship exists between the two.

	Controlled Entry Percentage	Entries/60	wE/60
Fenwick For/60 (On-Ice Shot Creation)	0.3103	0.1162	0.2501
Fenwick Against/60 (On-Ice Shot Suppression)	0.0072	0.0076	0.0005
Fenwick For Percentage (On-Ice Shot Differential	0.1372	0.0923	0.1497

There a few obvious findings here. First, the strongest relationship is between the entry metrics and shot generation metrics, which makes perfect sense because entry creation is rightfully viewed as an offensive skill. There is no obvious relationship between entries and shot suppression, implying that players who are adept at making plays with the puck in the neutral zone are not necessarily stellar in terms of shot suppression. This finding makes sense in light of Alex Novet’s previous research regarding Entry Generation and Suppression, and Micah Blake McCurdy’s finding that shot generation and shot suppression are mostly independent.

There is a correlation between entry generation and on-ice shot differential for individuals, but this is almost certainly driven by the offensive boost it provides for the team. Poor individual entry generation likely can hurt a player’s Corsi or Fenwick, but the issue would be isolated to the shot generation side of the equation.

Another finding is that, yet again, the Entries/60 metric lags behind both CE% and wE/60, just as it did in general repeatability. The gap is most obvious when looking at the shot generation R2s, where wE/60’s relationship is over twice as strong and CE% is nearly three times it.

Now, let’s move onto the defensemen in our dataset.

	Controlled Entry Percentage	Entries/60	wE/60
Fenwick For/60 (On-Ice Shot Creation)	0.0542	0.0800	0.0966
Fenwick Against/60 (On-Ice Shot Suppression)	0.0003	0.0176	0.0060
Fenwick For Percentage (On-Ice Shot Differential	0.0168	0.0671	0.0570

The general findings from the forwards hold true for the defensemen, as well. Yet again, shot generation and entry generation have the strongest relationship, with on-ice shot suppression barely moving the needle. However, the relationships on the whole are much weaker than those of the forwards. This is likely due to the fact that defensemen simply generate less entries than forwards on average. During the 2013-14 season, Erik Karlsson’s 7.14 wE/60 easily led all defensemen; that would have ranked him 386^th among forwards, just behind Steve Ott and Taylor Pyatt. Since their individual entries have less of an impact on the game when it comes to raw volume, it makes sense that defensemen wouldn’t move the needle much in this regard when it comes to shots, at least compared to forwards.

Out-Of-Sample Predictive Testing

We now know that when looking at individual player entry generation metrics, they have their most value in terms of their connection to on-ice shot generation, particularly for forwards. We also know that success in terms of entry creation is repeatable on the individual player level. What we still don’t know, however, is just how important these metrics are when trying to project future results.

After all, when you do an in-sample test of PDO versus single-game winning percentage, the correlation is extremely high. But when you try to use that metric to predict future single-game winning percentage, that’s where the cracks in its usefulness begin to show.

To account for this, let’s check to see if entry generation in odd numbered games from 2013-14 predicts strong on-ice shot generation in even numbered games. Since we determined that individual entry metrics are most useful for forwards, we’ll limit our study to that position. This method will also allow us to pit the three metrics against each other, and better understand which truly are most valuable in judging entry generation talent.

Entry Metric (in odd-numbered 2013-14 games)	Relationship with FF60 from even-numbered 2013-14 games
Controlled Entry Percentage	0.2178
Entries/60	0.0655
Weighted Entries/60	0.1602

Controlled Entry Percentage leads the way. And though it falls short of Controlled Entry Percentage, wE/60 performs dramatically better than its unweighted counterpart Entries/60. In repeatability, in-sample testing, and now out-of-sample testing, wE/60 proves to be the superior metric in judging entry volume creation.

There remains one more test that we can run on these entry generation metrics: whether they show any value in predicting on-ice goal generation for forwards. We’ll also add Fenwick For per 60 to the mix as well, to determine whether entry metrics can match shot metrics in their predictive abilities.

Metric	In-Sample Relationship vs. Goals For Per 60	Out-Of-Sample Relationship vs. Goals For Per 60
Controlled Entry Percentage	0.3032	0.2320
Entries/60	0.0872	0.0573
Weighted Entries/60	0.2217	0.1560
Fenwick For Per 60	0.3827	0.1908

We’re only dealing with one season here. But in this dataset, Controlled Entry Percentage actually outperformed FF60 in its out-of-sample relationship with Goals For Per 60, and wE/60 was within striking distance. This finding will be certainly be worth attempting to replicate once the complete 2016-17 tracking project is available for study.

Findings and Conclusions

We’ve long known that team-based entry statistics have value in measuring performance. Therefore, it should come as no surprise that there is also some value in using them on the individual player level as well. But as with any new metric, there comes a temptation to overreach and overrate its usefulness, which is where testing provides its true value, resulting in provable conclusions regarding best practices. Below are those findings.

• Entry creation appears to be repeatable on the individual player level.
• Entry metrics are best viewed as a component of on-ice shot generation. The relationship between entry creation and shot prevention is nonexistent.
• Weighted Entries/60 (wE/60) should replace raw Entries/60 as the best way to measure individual entry creation volume. It is more repeatable, correlates stronger with on-ice shot generation, and predicts both shot and goal generation better than E/60.
• Individual entry metrics are more useful in evaluating forwards than defensemen.
• If only one entry metric is available, Controlled Entry Percentage appears to be the most valuable in evaluating forwards. It showed the highest level of repeatability, was correlated strongest with on-ice shot creation, and predicted both shot and goal generation better than wE/60.
• While Controlled Entry Percentage appears to be superior to wE/60 in terms of presenting the degree to which a player’s neutral zone performance affects his team’s shot generation, wE/60 still retains value because it directly measures the total individual contribution to expected shots via neutral zone play. Also, there are players in this dataset (Tyler Bozak, Alex Tanguay, Thomas Vanek) who combine high Controlled Entry Rates with low entry volume. Identifying those individuals likely would help in structuring effective line combinations, since stacking a line with too many forwards who look to carry the puck through the neutral zone only in the most idyllic of circumstances risks negatively impacting overall shot generation.

This research comes with the caveat that we are only looking at one full-season worth of data (2013-14). With a larger dataset, more concrete conclusions can be made regarding the usefulness of these metrics, particularly in terms of the predictive power of Controlled Entry Percentage and wE/60. In addition, utilizing entry metrics as a component of larger studies (such as Ryan Stimson’s work on Clustering and Playing Styles) should serve to extract full value from these manual tracking projects. Ryan’s work on the impact of a successful completed pass following an entry upon shot and goal generation is yet another example of the value of “interzonal” research.

Still, it’s helpful to know how to utilize individual player entry metrics in isolation. If a player’s on-ice shot creation metrics start to decline, a look into whether his CE% or wE/60 is showing signs of decline as well could help to explain the dropoff, and provide coaches with a point of emphasis to address with the player. By the same token, if a player’s offensive creation seems to be in decline but his entry metrics remain strong as ever, that can hint that the issue is either random variance or is confined to the offensive zone (and not the neutral zone). Ruling out potential causes of a decline is an essential part of the process of adequately addressing it.

In any case, manually-tracked metrics like Controlled Entry Percentage and wE/60 clearly have their place in the statistical community, so long as they are used within the proper context and parameters. And with Corey’s newest dataset nearing completion, we should soon gain an even better understanding of just that.

All metrics from Corey Sznajder’s 2013-14 All Three Zones project, and from Corsica.Hockey.