A little more than a year ago, the NHL announced that it had officially adopted a 3-on-3 overtime as a means of juicing goal-scoring in the league and avoiding the shootout to as great an extent as possible.
At the time, we had little understanding of what the NHL looked like at 3-on-3. From 2007 through the end of 2014-15, all NHL teams combined had played a grand combined total of just 87 minutes in that strength situation. This is a situation most teams never practiced, because it was so rare that it wasn’t worth anyone’s time to actually come up with systems or waste precious time drilling for the maybe 30 seconds a year they’d spend playing 3-on-3 hockey if they were quite lucky.
Based on the evidence available at the time, though, there wasn’t a lot of reason to believe that teams would score so much more often that it really moved the needle. A rough estimate based on goals per five minutes at 4-on-4 versus 3-on-3 showed that the league might cut shootout occurrences to once every nine games from the previous 1-in-7.5 or so.
In actual practice, the League started out hot and cooled down — as many predicted it would — potentially because teams had “figured out” how to coach against even wide-open 3-on-3 tactics. The NHL, as it so often does, ground the fun out of even the most fun thing it could produce in the regular season. One number-crunching effort from USA Today showed that in the first three months of last season, only a little more than 1-in-3 overtimes went to a shootout. Once the new year rolled around, that number grew to nearly 1-in-2. You can explain some of that away as variance; sometimes pucks just don’t go in despite everyone’s best efforts. And still, the league was deciding more games with the shootout than not (60 percent through mid-March versus 42 percent the year before).
But the question is, mathematically speaking, whether the league actually figured out how to reduce the odds of either team scoring or whether goal-scoring went away by chance.
The latter is an entirely possible cause for the apparent decline in overtime goal-scoring.
To understand that, though, we have to get a baseline of what the league looked like after a full year of 3-on-3 overtime versus what the same league looked like the season prior at 4-on-4. As you might imagine, the number of events increases somewhat drastically on a per-5-minutes basis when you get one extra guy per team off the ice:
So what can we discern from those numbers? Well, the first is that a greater percentage of attempts are actually getting through to the net, and given that goal-scoring exploded on a per-five-minutes basis – thanks to both a dramatic increase in shots on goal and shooting percentages – we can also surmise that the quality of shots in general increased significantly.
(Sad to say, but the loss of War On Ice leaves us wanting for information on high-quality chance volume for now; Corsica’s Rink View feature remains a work in progress.)
That all stands to reason. More space on the ice, and a greater distribution of overtime TOI to higher-quality players — i.e. you rely on your six-best players instead of your eight-best — means more scoring in these situations. This is to say nothing of the advantage that comes with 4-on-3 power plays that became somewhat common in OTs this year.
But again, these are baseline numbers. The League had years to figure out how best to operate at 4-on-4 and had whittled it down to an unsatisfying science by the time 2014-15 rolled around. Surely, the 3-on-3 numbers observed in, say, October and November would likewise dwarf those in March and April.
To determine whether that was the case, I looked at the season in roughly two-week increments: the 1st-through-15th of every month, then the 16th to the end. Obviously this created results that were a little skewed, because the most demonstrably skilled teams tended to be the best when it came to generating expected-goals over the course of the season, and they didn’t reliably play the same number of minutes per two-week period all year long. Teams like Chicago and Pittsburgh sometimes played two or three overtimes in 15 days, and sometimes they played none. Meanwhile, Nashville and Anaheim, which generated the fewest expected goals per five minutes at 3-on-3 this past season, were in the same boat.
But honestly, when you look at the numbers, there is a relatively small downward trend in terms of things like shot attempts and even shots on goal over the course of the year.
Most noteworthy, though, is that the USA Today numbers about the state of OT before and after Jan. 1 hold up to some extent. Teams regularly combined to score 2.2 or so goals per five minutes of overtime in the first three months of the season, and by late December it just dropped off a cliff, with goal-scoring falling about 40 percent on average.
Teams did get better at limiting shots on goal and attempts, but most notable is that the five two-week periods from the start of October to Dec. 15 saw an average shooting percentage of about 15.7 percent. From Dec. 16 through the end of the regular season on April 10, that number was a shade under 12 percent. It’s a substantial enough drop that the League should be worried, especially because it’s accompanied by fewer shots (about 1.25 fewer per 5 minutes) and attempts (down 1.3 per). That means that pretty much the entirety of of the drop-off in events came because teams were just getting better at taking shots away.
I think there’s probably still room for teams to get even more conservative in 3-on-3 situations, but at some point there’s a logical end to how much even the most defensive-minded coaches can limit offense in such situations. It’s hard to say what the “bottom” here is, but we might be close to it right now. The more overtime we see, the more data we get, and the better off we are when it comes to determining how effectively the league is dealing with this issue.
Remember, we only had 87 minutes of 3-on-3 from 2007-15, less than 11 minutes per season. This past year we saw nearly 797, about 53 minutes per team. Some, like St. Louis (more than 80 minutes) played significantly more than that, while others like Arizona (a bit over 30) managed far less. But as more coaches get more exposure time, they’ll probably chip away at 3-on-3 scoring. That, in turn, makes shootouts more common.
But even if things get as conservative as possible under the new 3-on-3 format, it’ll probably still be a lot better than 4-on-4.
All stats via Corsica unless otherwise stated.