About a month ago, while staring at strong possession teams and trying to figure out why some had counter-intuitively poor zone start numbers, I struck upon a remarkable coincidence between zone starts and icings. I haven’t had a chance to more fully explore this coincidence, so this piece will basically serve as a hub for all the information I can gather on the topic in the hopes of discovering more answers in the future.
The Elements
1. Possession
Let’s establish a graph of the 30 NHL teams ordered according to possession (here, I’ll use Fenwick For Close Percentage):
Now, my expectation is that a team’s zone starts (the percentage of a team’s faceoffs taken in the offensive zone) would correlate highly with its possession numbers. The idea being, a team that drives possession (i.e., takes a higher percentage of shot attempts) would create more offensive zone faceoff opportunities. A team that always has the puck and generates a lot of shot attempts seems likely to start the majority of its shifts in the offensive zone.
2. Zone Starts
You can probably note a few outliers making a mess of my assumptions. But, let’s put these graphs side-by-side to bring the point home:
If we create three sets here and define their criteria as +/- 1-5 spots = “within range”; +/- 6-10 spots = “out of range”; and, +/- 11-x = “well out of range,” we find that the vast majority of the teams fall “within range” (20 teams), leaving a third of the teams “out of range” (4) or “well out of range” (6).
Two thirds of the NHL, then, suggest there is at the very least a strong correlation (if not a causal link) between possession and zone starts. And, yet…
The exception is more interesting than the rule. The rule proves nothing; the exception proves everything: It confirms not only the rule but also its existence, which derives only from the exception. In the exception the power of real life breaks through the crust of a mechanism that has become torpid by repetition.
~Carl Schmitt, Political Theology
The exceptions, as they always are, are the interesting ones. In the above graph, I’ve highlighted those teams that are defined as “well out of range.” These are teams that have a possession to zone start differential of +/- 11 or greater. The obvious question prompted here: how does a great possession team like the San Jose Sharks manage to enjoy relatively poor zone start numbers?
[One fleeting idea: perhaps there is a difference between possession and shots (i.e., traditional shots on goal) that explains the zone start differential? The idea being: shots on goal have a higher chance of leading to a faceoff situation than shot attempts (which includes shots that miss the net) and perhaps a team like the Sharks, while generating a lot of shot attempts somehow manages to both miss the net a lot and maintain possession reducing their chances at offensive zone faceoffs. This idea stuck me as wildly implausible, but I checked it out anyway. There’s no there there as they say: the Sharks are 3rd in the league in shot percentage, 3rd in total shots for and 2nd in shots for per 60 minutes (all numbers 5×5 close).]One obvious place to start looking, considering we are dealing with faceoff situations in zone starts, was icings.
3. Icings
Before going dark, extraskater ran a table of the 30 NHL teams by icing drawn percentage on the blog (which, hint hint, is still available).
As with zone starts, a reasonable assumption would be that strong possession teams (teams that always have the puck in the offensive end) will draw a lot of icings as they will force the opposition into taking sub-optimal paths to relieve pressure.
[Caveats: with icings we are dealing with a relatively small sample size and we need to be careful in how we interpret these numbers. Moreover, we are here dealing with “drawn percentage” and need to be mindful that it is possible for a team to be very good (or bad) at drawing icings and yet have a poor icing drawn percentage because they also happen to be very poor (or good) at taking icings. For example, Anaheim and Columbus have less than a percentage point separating their icing drawn percentage (49.5 and 48.7 respectively) and yet Anaheim managed to draw 102 more icings during the season, which amounts to 27% of Anaheim’s total icings. For the raw icing numbers see extraskater’s blog post.]If we use the same sets outlined above (“within range,” “out of range” and “well out of range”) and compare possession numbers with icing drawn, we see more volatility than we did with zone starts, but there is still a fair amount of correlation.
13 teams fall “within range,” 9 “out of range” and 8 “well out of range.” Of the 6 teams we originally identified as being “well out of range” regarding their possession and zone start numbers (Sharks, Canucks, Jackets, Predators, Flyers and the Hurricanes), 4 teams (Sharks, Canucks, Predators and Hurricanes) remain “well out of range” with the other two (Jackets and Flyers) nipping at their heels and “out of range.”
The ultimate idea here is to try and establish whether icing drawn percentage can explain, in part at least, the zone start numbers we are seeing respecting the “well out of range” possession-to-zone start teams.
Looking at the San Jose Sharks, for example, we find that the 3rd best possession team in the league is 20th in zone start percentage. Can icing drawn percentage explain part of the gap? Considering the Sharks were 27th in the league in drawing icings, it certainly seems plausible.
In Vancouver, we see the same phenomenon. 8th best possession team. 25th in zone starts and 29th in icings.
In Carolina, we see the same phenomenon reversed. 24th best possession team. 8th in zone starts and 7th in icings.
The same pattern is roughly repeated by Blue Jackets, Predators and Flyers.
There are a few other extreme cases of possession vs. icing disparity that perhaps explain something of the possession-zone start differential. In the cases of the Capitals, Stars, Jets and Avalanche we can see that the icing numbers appear to be pulling the zone starts away from the possession numbers.
Dallas is an example of a good possession team that appears to have its zone starts negatively affected by its icings. 9th best possession team. 13th in zone starts and 22nd in icings.
The Capitals, Jets and Avalanche all represent the opposite: poor possession teams seeing their zone starts positively affected by icings.
Capitals 25th possession team. 17th in zone starts and 2nd in icings.
Jets 18th possession team. 11th in zone starts and 6th in icings.
Avalanche 27th possession team. 22nd in zone starts and 8th in icings.
Based on these exceptions (teams with zone starts out of sorts with their possession numbers, which see some correlation between their zone starts and icings), it seems reasonable to suggest icings are a non-trivial contributing factor affecting a team’s zone start percentage.
Here’s a table putting all three elements together:
Literature Review and Analysis
It is impossible to get a handle on all the hockey content out there. But, from my own research, I’ve found that little has been written about icings in relation to zone starts and/or possession. Ryan Pike has suggested that icing differential may, at both the team and player level, be an important tool for evaluation. However, his article operates as more of a heuristic than an analysis. The one study that comes closest to my own was done by Gabriel Desjardins back in the 2009-10 season. At the time, he noted, “icing, which we normally associate with a weak defensive team desperately trying to clear the puck – does not seem to correlate very well with that image.” He followed this with a table showing the league leaders in icing drawn percentage*:
He rightly noted that Edmonton (a weak team then, as always) led the league in this department, in essence flouting common sense. Though he didn’t do a direct comparison of these icing numbers with possession numbers from the 2009-10 season, we can do so now.
The results are actually quite similar to what we found comparing the 2013-14 possession numbers to icing percentages. 13 teams are “within range,” 7 “out of range” and 10 are “well out of range.” It certainly appears that icing percentage doesn’t have a strong effect on possession. However, that isn’t quite what we are interested in here.
We are interested in seeing if icing percentage can explain some part of how good or bad possession teams manage to also have unexpected zone starts, i.e., how a good possession team might have a relatively bad zone start percentage.
Using Desjardin’s icing data for the 2009-10, let’s run the exercise again.
Unlike what we saw in the 2013-14 season, here the differential between possession and zone starts is far less striking. Only one team (Atlanta) manages to be “well out of range.” A full 18 are “within range” and 11 are “out of range.” In order to test the hypothesis here, I’ve expanded my range to +/- 9 (from +/-11). This gives us a sample of 6 teams.
As you can see, with the exception of Phoenix, all the teams appear to have their zone starts pulled up or down by their icing numbers in relation to their possession numbers.
Nashville, the 6th best possession team is 16th in zone starts and 28th in icings.
St. Louis, the 11th best possession team is 20 in zone starts and 23 in icings. etc.
While this is hardly conclusive, I believe the pattern here suggests some kind of correlation between a team’s ability to force icings and their zone starts. While a team’s icing percentage may not have a strong impact on their possession numbers, it looks like it may have an impact on where they start each shift.
Avoid Using Ice Related Puns
The crucial question left dangling here is: what are these exceptions, these teams with zone starts “well out of range” of their possession numbers, doing that causes them to take a high percentage of icings (or, conversely, what are they doing that forces the other teams to take a high percentage of icings)?
My assumption is that these teams have systems in place for various on-ice situations, which is leading to these positive or negative effects. To confirm this, we’d have to do extensive video review and make some educated guesses.
In the meantime, an interesting note cropped up in Scott Cullen’s preview of Corey Sznajder‘s immense project of tracking every single NHL game from the 2013-14 season. Cullen notes:
One of the surprises he has found is that there is a team that is a great possession team, the San Jose Sharks, that isn’t particularly strong in the neutral zone. That’s somewhat unexpected.
Perhaps, in the case of the Sharks at least, this is the answer. Perhaps there is something about their system that leaves them vulnerable in the neutral zone and prone to icing the puck and having relatively poor zone starts.
[adsanity id=1808 align=alignnone /]Add The Sports Daily to your Google News Feed!
