Measuring the real impact of NHL’s new 3-on-3 overtime (Trending Topics)

Washington Capitals' Troy Brouwer scores past Montreal Canadiens goalie Carey Price in the shootout of an NHL hockey game Thursday, April 2, 2015, in Montreal. Washington won 5-4. (AP Photo/The Canadian Press, Paul Chiasson)

Earlier this week, the NHL announced that its five minutes of 4-on-4 overtime would, instead, be replaced by five minutes of 3-on-3 for the coming season as a means of reducing the number of shootouts.

This is a good idea because the shootout is bad, and it stands to reason that more goals will be scored at 3-on-3 than 4-on-4. Remember, the AHL went to a sort of hybrid format for this past season: three minutes of 4-on-4 before another four of 3-on-3, then the shootout if nothing was decided.

And with that system — which the NHL did not fully adopt — 75 percent of overtime games were decided before the shootout became an option. That was more than double the previous season. So effectively, that shows how effective this can be in getting rid of the shootout which, again, is stupid and bad.

And since the shootout was implemented, the share of NHL games that went to overtime has always been between about 23 and 25 percent. In an 82-game season, that amounts to about 290-300 games, or roughly 10 loser points per team per season (though obviously not distributed evenly). And of those 290-300 games, it's usually about 50-60 percent that are decided in a shootout. Over a 10-season period, this is what it looks like:


So basically, everyone agrees 13 percent of all games (160 per year) is too great a portion to basically have games decided on a skills competition. Everyone is also right on that count. If the league can get that number down to the AHL's no-shootout success rate — I'm dubious of that for reasons we'll discuss in a second — that means only about 6 percent of all NHL games would be decided with a shootout. A little more than 1 in 16, which is a lot better than 2 in 15. Assuming that next year is a perfectly average season in terms of the number of games going to the shootout, this is about what the breakdown would look like:


If you're wondering, that means that only about 74 games out of the league's 1,230 next season would go to a shootout. You'd much rather have that than the current 158 or so.

But the reason I doubt that the league will have that level of success in reducing shootout instances is that, frankly, three minutes of 4-on-4 then four of 3-on-3 is going to give you a lot more goals than just five minutes of 3-on-3. The extra two minutes per game may not seem like a lot, but in terms of the impact it would have on goal-scoring, it's actually pretty big.


The first thing to keep in mind about the event rates for 3-on-3 hockey is that it's incredibly rare, and therefore subject to huge swings in a short amount of time. Over the past several years, in fact, teams have played a little more than 87 minutes at 3-on-3 (scoring nine goals in that time). Total. In the entire NHL. The other thing to keep in mind is that, because of this, no one really practices 3-on-3 hockey and so it's not much of a surprise that rates for everything but goals take a bit of a step back.

However, when you dig into the numbers a bit more, you find that while the number of shots on goal per shot attempt holts steady at around 56.8 (I use this as a kind of shorthand for shot quality; blocks obviously go way down with fewer people on the ice), The conversion rates on high-quality chances increase greatly. While it can't be reflected in the data, the eye test and common sense dictate that there would naturally be more odd-man rushes and breakaways at 3-on-3 than 4-on-4 or, certainly, 5-on-5. Those numbers are also reflected in shooting and save percentages.


The number that's most interesting there is, obviously, the fact that save percentages go down 20 points. Again, you're dealing with volatile numbers when you get down to only about 87 minutes of hockey, and I wouldn't normally expect a decline to be that sharp over a few thousand minutes. But the drop would probably still be fairly big, and that's why the AHL was only going to shootouts in 6 percent of its games as a result.

Moreover, because teams will soon be practicing 3-on-3 a hell of a lot more than 4-on-4 given these rule changes, I'd expect better quality of play and systems in those situations in general. That, in turn, probably boosts rates for attempts, shots, and high-quality chances at the very least. Goal numbers, then, end up taking care of themselves.

But the question becomes if the NHL can reasonably expect to avoid as many shootouts as the AHL — even given the higher skill level at the top level of the game — with a straight five-minute 3-on-3 OT versus the seven-minute mixed format. Using the numbers we have above, we can approximate success rates for so short a timeframe in each overtime game. And the answer is, “Nope.”


The number of shot attempts and shots on goal under an AHL format go up almost 43 percent each, high-quality chances increase 45 percent, but most crucially, goals rise about 31 percent. Basically what that last number means is the NHL's policy will probably reduce the number of overtimes going into a shootout, yes, but probably nowhere near as much as people seem to think right now. There would, obviously, be even sharper increases in goal-scoring if it were seven minutes of 3-on-3 instead of the AHL former. (It was the NHLPA, though, that balked at seven minutes of OT; who can blame them if they're getting paid the same?)

Again, right now about 57 percent of all overtime games go to a shootout. The new NHL 3-on-3 format, if these numbers hold up (though they probably won't), would reduce that to about 49 percent. Adopting an AHL 4-on-4-then-3-on-3 format would cut it to 33 percent or so. Going with my theoretical hybrid of the two would slash that to 28 percent.

That, at least, begins to approach the AHL number, which I am willing to bet is a result of variance rather than a hard-and-fast number we can expect every year from here on out. And please keep in mind that I'd expect the NHL numbers to change dramatically when we start dumping hundreds of minutes of this stuff into mix over the course of the season.

Mathematically, we can expect the 87 or so minutes of 3v3 time in the NHL seen since 2007 to have doubled by late October or so (if you assume the since-2005 average of 4.2 minutes of overtime per night of the NHL's 180 nights in a season, or 87 minutes approximately every 21 days).

But basically, the takeaway here is that the NHL currently sees a shootout once in every 7.5 games (13 percent). The new format will reduce that number to a little less than once per nine games (11.7 percent).

Meanwhile, though some have said it's likely to be a boon for goalscoring rates, if you're only adding 16 goals to the entire NHL total, it's a 0.24 percent increase. Negligible doesn't begin to describe the impact. If a player is extremely lucky and/or good at 3-on-3, he might wring an extra two goals out of this per season. Wow.

In all, this amounts to about 144 bonus points for winning in a shootout, down from the current 160ish. That means that, on average, each team will lose about half a point in the standings as a result of shootouts over the course of a season.

That's progress. But really, is it enough progress? You'd have to say no.

If anything, this looks to me like someone said, “Well 3-on-3 is going to give you more goals than 4-on-4,” without doing the math to support the hypothesis. And while they're technically right, the reduction in shootouts is going to be miniscule.

For people who want the shootout gone, these changes don't really show a dedication to that. They show a kind of lazy willingness to appear as though they're dedicated. But that's about it.

(All data via War on Ice)

Ryan Lambert is a Puck Daddy columnist. His email is here and his Twitter is here.