Blue Jackets head coach and general gift to the media John Tortorella made an interesting point Friday morning after Columbus beat Philadelphia in overtime on Thursday.
“Get rid of the shootout and just play the 3-on-3 until a team dies. I think that’s the way it should be. It’s exciting!… You know I’m not a gimmick guy, I think it has to be determined within the game. Three on three, it still isn’t a total team but it’s still 3-on-3. Let the games end there. I think it’s exciting. I have to watch myself. I get caught up watching and I almost forget who’s up next because I’m watching the 3-on-3, so I think it’s really good for the game.
Advertisement
3-on-3 until you die. It’s gonna be over. It’s gonna be over. I know there’s other things, TV, how much time the game…it’s not gonna last long, I’ll tell you right now. So, let ’em play.”
I’m very much in agreement with Tortorella on this point. It’s fun to watch 3-on-3, more fun than it is watching a shootout. He alludes to the critical question though: How much 3-on-3 would you need to guarantee a winner?
So far this year, there have been 1,313.3 minutes of overtime played, with 158 goals scored. That’s 14.4 goals per 60 minutes. Just for contrast, in the first sixty minutes of the game, there have been 5.9 goals per 60 minutes. So yeah, goals spike in overtime, most of which is 3-on-3.
There’s a math trick that you can use to explore this question. If you know the rate at which goals are scored in overtime, you can predict the likelihood of a game ending within a certain period of time. There have been 220 overtime games played through Saturday night. I graphed the expected likelihood of a game ending by a certain point in overtime against what we’ve observed this year. As you can see, it lines up reasonably well.
I have a few observations about this. First of all, it’s important to keep in mind that I’ve used a constant for the likelihood of a goal being scored at any particular second. While that’s a nice, simple way of doing it, it’s not quite right. The most obvious example of this is the first few seconds after overtime starts. There’s a faceoff, the faceoff is won, and whichever team won it has three opposition players lined up in front of them. If we knew, for example, the rate of 2-on-1s and breakaways, I’d expect that we’d see it be very low in the immediate seconds after overtime begins, rising as the game shifts into the back-and-forth that defines 3-on-3.
Advertisement
There’s also a bit of a spike in games finishing toward the end of overtime. Through 184 seconds, the math trick I’ve used predicts 105.1 of the 220 games would still be ongoing. In fact, 112 games were still ongoing. By the end of overtime, the math predicts that 66.1 games will be ongoing; in fact, only 62 games have gone to shootouts through Saturday night. Three possibilities come to mind here. First, it’s nothing, a weird fluke. Second, coaches are managing their bench to get their preferred overtime guys on the ice as time dwindles. Third, we’re more likely to see penalties as overtime moves along and less likely to see penalties entering overtime.
I was curious, so I grabbed the data for the most used forward on each team in overtime and took a look at what percentage of the ice time in each 30 second overtime window was played by the most used forward on each team. You can see a definite period in the 61-150 second window where the most commonly used forwards are less likely to be on the ice and then it bounces back a bit toward the end of overtime.
So to the extent that the coach prefers the guys who are most likely to score, there’s probably something there. There’s definitely more special teams time as overtime progresses. Eighty-nine percent of the first thirty seconds of overtime were played at 3-on-3 with the goalies in. By the final 30 seconds, just 74 percent of time is being played at 3-on-3 with the goalies in. The share of ice time consisting of 3-on-3 minutes drops pretty consistently.
This probably tells us something about how overtime works and would be something worth considering when thinking about how longer overtime might play out. One of the things that might confound the analysis based on the existing data is that we might see more power play time in minutes six through ten of overtime than we see in minutes one through five. It’s hard to defend with three guys and eventually, someone has to make a choice between allowing a chance or taking a penalty. It does seem to flatten out a bit in the final 90 seconds of overtime but it’s worth bearing in mind that the natural level of special teams versus even strength time in 3-on-3 hockey might be something like a 75/25 split.
So what does all of this mean? Well, given that our method of estimating how many games would wrap up matches up reasonably well with what we’ve actually observed, we can estimate what longer overtime would give us over the course of the entire season. We’re on pace for 278 overtime games this year. Applying our method of estimating how many games will still be going on a certain point in time, we can look at how playing a longer overtime might impact things.
Our dumb model predicts that we’ll end up with about 83 shootouts this year. You can see the problem with Tortorella’s idea here: If the NHL played until there was an overtime goal, we’d expect to still have a few games that didn’t have a winner after 20 minutes. Even leaving aside the question of TV windows and travel and fatigue on the players, there’s the question of the playing surface. Unlike basketball, baseball and (mostly) professional football, hockey’s different in that the playing surface is degrading as time rolls along.
Advertisement
Really, to me, the league has a couple of objectives to balance here. They’re clearly of the view that it’s important that there’s a winner, that ties don’t fly in North America. At the same time, I think they’ve also indicated that they want to limit the extent to which shootouts are dictating the outcome of playoff races. We can see this in the decision to exclude shootout wins from the tiebreaker, as well as the decision to go to 3-on-3 overtime when it became clear that 4-on-4 overtime was still resulting in a lot of shootouts. In the last year of 4-on-4 overtime, 56 percent of overtime games went to the shootout. Since the change, it’s been 39 percent, 32 percent, 34 percent and, this year, 28 percent. It’s kind of interesting that there still seems to be an ongoing decline in shootouts.
The question, I think, is how you balance the interest in having a winner in a particular game with a desire to reduce the impact of the shootout. I’m not sure that Tortorella’s “play until a team dies” plan is really necessary, even though I agree with him that I’d rather things be decided at 3-on-3 than in a shootout. Looking at our simple model, here’s how many shootouts we’d expect this year with varying overtime lengths.
- 5 minutes: 83 shootouts
- 6 minutes: 66 shootouts
- 7 minutes: 52 shootouts
- 8 minutes: 41 shootouts
- 9 minutes: 32 shootouts
- 10 minutes: 25 shootouts
So you can see the diminishing returns a little bit. Adding an extra minute would be expected to knock out 17 shootouts. Adding an extra minute to go from nine to ten minutes knocks out less than half of that. If your objective is to diminish the value of the shootout in determining playoff teams to the point that it’s practically irrelevant, you probably don’t need the “Play Until Death” plan. A few more minutes would probably take care of it and you probably wouldn’t even need to get to ten minutes. Even adding three more minutes of overtime would be expected to halve the number of shootouts, subject to some of the issues I’ve flagged above about special teams (maybe it would wipe out even more shootouts with minutes six through eight being more likely to feature power play time!) and who plays the minutes (maybe teams would have to use lesser players more as overtime rolled along, reducing the level of offence.)
It’s not a perfect solution but, if the ice and TV issues could be resolved, it’s a change I’d endorse. We’d be guaranteed a winner every night but we’d also all but wipe out the impact of the shootout on making the playoffs. There’s a balance to find there. Truthfully, the league is probably already pretty close to it. Looking at the three years in which the league has played 3-on-3, the closest thing I can find to an argument that the shootout swung a playoff race comes in 2016-17, when the Senators went 6-4 in the shootout while finishing four points ahead of the Islanders, who missed the playoffs with a 2-4 shootout record and 94 points. It’s a pretty marginal case.
So, while I agree with Tortorella’s overall point, you can see the reasons why it would be silly for the league to go to a rule whereby 3-on-3 overtime was played until there was a winner. They don’t really need to though, in order to basically wipe out the impact of the shootout on playoff qualification. A few more minutes of overtime would probably get them there. And as a bonus, be pretty awesome to watch.
(Top photo credit: Aaron Doster-USA TODAY Sports)