One of the central battlegrounds between the fancy-stat set and hockey traditionalists is “puck possession.” Most practitioners of hockey analytics believe that puck possession – as measured indirectly by stats that look at shot attempt differential – are powerful predictors of performance. The traditionalists tend to focus more on such things as experience, grit, and other “intangibles”.
This year, however, has been a big year for demonstrating the value of puck possession.
Indeed, the value of possession was very much under the microscope in hockey’s biggest market, as the Leafs tried to put together a successful season despite playing almost all of every game in their own end. They succeeded for a while – 66 games, to be exact – but, in the end it turned out to be too much for them, and Leafs haters were treated to their third dramatic collapse in as many seasons.
This year’s playoffs are also shaping up to be a lesson in the value of possession. Twelve of the top sixteen possession teams made the playoffs, and of the four teams left, three are who you would expect based on their possession stats in the regular season: L.A. was 1st overall in shot attempt differential, Chicago was 2nd, and the Rangers were 6th.
None of the four teams would have been odds-on favourites to make it this far, if you just looked at the standings.
The power of possession is illustrated even further by examining how individual series have played out. So far, based on the standings, there have been 7 upsets in the 12 series played. But based on possession metrics, only 4 times has the team with the worse possession stats in the regular season won. Every team left has beaten somebody ahead of them in the standings, but only Montreal has had to beat a team that put up better regular season possessionnumbers, and they did that twice. Interestingly enough, they did that by dominating the possession game in the playoff series.
This year’s playoffs aren’t unusual. Using data from David Johnson’s site, Hockey Analysis, we looked at the probability of winning a playoff series as a function of the differential in the teams’ (a) points in the regular season, and (b) shot attempts. Since possession data only goes back to the 2007-08 season, this meant we had 102 series to draw upon, including the first 2 rounds from this year.
What we found was quite interesting.
First, both the regular season standings and the possession metrics factor into the probability of winning. Each additional point a team had over its opponent in the regular season translates into an increased probability of winning the series of 0.6 percentage points. This is not a particularly big effect, as it means that a team who accumulated 10 more points in the standings is only a 56-44 favourite.
For possession, an advantage of one percentage point during the regular season translates into an increase in the probability of winning of 2.5 percentage points.
Although this makes it seem like possession differential is much stronger than point differential, in practical terms, the predictive power is about the same because the spread between two teams’ possession stats is usually much less than the spread of their regular season points. To give some context, the biggest difference we saw in this year’s playoffs in terms of points was in the Anaheim-Dallas series. The Ducks amassed 25 more points during the regular season, which translates to about a 15 percentage point advantage. The biggest difference in terms of possession, however, was in the Boston-Montreal series. The Bruins were 5.7 percentage points better in terms of possession, which translates to an increased probability of winning of about 14.25 percentage points. Given the 17 point advantage Boston held in the standings, this was a significant upset.
So what does this mean? First, it underscores the importance of puck possession in the NHL. But, perhaps more importantly it gives a bit of the lie to the saying that teams are as good as the standings say they are.
| Team | Points | Rank in Conference | Possession (Fenwick For %) | Rank in Conference |
| Blackhawks | 107 | 5th | 55.2% | 2nd |
| Kings | 100 | 6th | 56.7% | 1st |
| Canadiens | 100 | 4th | 48.4% | 11th |
| Rangers | 96 | 5th | 53.6% | 3rd |
Then how do you explain the New Jersey Devils 2013-14 playoff-less season?
Well, first of all, this study was looking at the predictiveness of regular season possession for success in the playoffs, not getting into them. So the fact that the Devils didn't make the playoffs is not very relevant to the analysis here. What it does suggest, however, is that if they had managed to sneak in as a wildcard, for example, then they would have a very good chance of pulling off an "upset".
And secondly, nobody is saying that possession stats perfectly correlate with success. There will always be teams that are good at possession who don't qualify for the playoffs, as happened to the Devils this year and last, and teams that get into the playoffs despite being poor possession teams. Possession isn't everything.
I don't disagree with the premise - that possession stats are important - but using 'regular season' possession stats to justify playoff outcomes is far less relevant than using playoff possession stats.
Take a look at the head to head playoff C% and you'll see that quite a few series were won by the teams with the better PDO (goaltending and shooting %) rather than the ones with the better C%.
Teams play different in the playoffs - you can't ignore that in your analysis.
Hi Kato,
Thanks for the comment. Your comments about the importance of goaltending and shooting are well taken, but they are important for answering a different question than the one I was looking at. Possession metrics within a series as well as PDO help explain why teams win series from a retrospective viewpoint. That is, it helps us understand the relative importance of these things in determining the winner of a series. The question I was more interested in was more predictive in nature. That is, before the series starts, which team should we view as the favourite and which is the underdog? And by how much? As it turns out, teams with fewer points in the standings may be more likely to win a series than the team ahead of them - if they were a better possession team. This is not at all obvious, as there already is a strong correlation between a team's possession stats and its regular season points. But it turns out that, even after a full season of play, possession stats matter above and beyond how they affect a team's points. One possible (and in my opinion, likely) explanation for this is what you've pointed out - teams play differently in the playoffs, and possession matters even more then.