About that late 5-on-3 against St. Louis…


I live in my parents’ basement, but sometimes I get to escape upstairs and watch a hockey game.

It’s weird to feel emotionally detached from the Vancouver Canucks, but after a year of having written about pretty much every team in the league, there’s no real worry for me to watch a game or not. More often than not I’ll watch when it’s convenient. It’s not that anything the organization has done that sucked all the hockey fan out of me, it’s that there’s just so much more I have to do, and especially on a Sunday night, I’d prefer to spend an evening reading than watching hockey.

But the Canucks happened to be playing the St. Louis Blues, a team that I find super interesting given how good they are with puck-possession and how bad they are in goal. Last season was odd, since Brian Elliott was all-world and he wasn’t supposed to be. This season he’s been poor, Jaroslav Halak has been hurt, and Jake Allen is carrying the reigns in the absence of both ‘tenders.

So I was watching the Canucks and Blues game, or at least the third period, cutting in and out of reading. St. Louis took a 3-2 lead with 12:00 to go in the period, putting the Canucks’ in a tight spot. I didn’t have my calculator handy, so I couldn’t peg the Canucks’ chances at a comeback. Besides, up until that point, they looked spectacular in the third period. They had out-shot St. Louis 6-0 in the period and had 13 Corsi events to St. Louis’ 4. I’m not sure what the scoring chances were at that point, but on shots and missed shots counted by the NHL play-by-play sheet, the Canucks had five attempts under 35 feet and St. Louis didn’t have a single one.

Knowing how good St. Louis is in puck-possession, I thought it was generally a good performance by the Canucks in that period even if they didn’t finish. This two-game losing streak aside, the Canucks have looked good in three of the four periods I’ve watched, again like a team that will walk over opponents once the goalies remembered they’re allowed to make saves. The scoring isn’t what it was in 2011 obviously, but there are better depth players on offence to cushion the Sedins’ slowing down in their advanced age.

All that aside, I was watching with my father. St. Louis scored the 3-2 goal 8:00 into the period. Vancouver got a powerplay 2:44 later, and with a boarding call a few seconds later on Alex Pietrangelo, that became a 5-on-3.

My father was quite happy, and he was sure that the Canucks were going to tie it up on the :41 second two-man advantage. I was less sure, having done the research a couple of weeks ago that teams generally strike at a rate of .79 per two minutes of 5-on-3 time. Doing some mental math, that the Canucks’ had about 33% of a regular 5-on-3 to work with, their chances of success were significantly lower: about 26%.

So I said “I bet they don’t score on this powerplay” quite confident in my prediction. Regardless of whether the Canucks are an above average 5-on-3 team (they aren’t) and regardless of the score, the data is quite clear that the length of a powerplay determines quite a bit. 5-on-3s are successful at a 28% rate, but that’s because the average 5-on-3 is :46 long. With less time to work with, the expected rate of success drops. Since the number I came up with was well below 50%, I was, again, very, very confident in my prediction.

Anyway, David Backes promptly won the ensuing face-off and sent the puck down the ice. I can’t remember exactly the progression of events, but if I recall correctly, the Canucks had a couple of botched zone entries the 5-on-3 was killed off without a shot or a scoring chance.

Still, I hadn’t anticipated the extra length of the powerplay. Even though the Canucks have generally been bad at 5-on-3s in the last couple of seasons compared to league average (the Canucks strike at a rate of .443 per two minutes, while the league hits at .739) the extra time afforded the club by the remaining 1:19 of the penalty after the 5-on-3 was killed should have factored into my prediction. Since the Canucks strike at a 5-on-4 rate higher than NHL average (.271 per two minutes to .199) you have to add the Canucks’ expectation over 1:19 to the 5-on-3 expectation.

Totalled up, once the Canucks got the extra man on the ice for a 5-on-3, they had a .151 chance of striking on the 5-on-3, and add to that an addition .178 chance of striking on the remaining 1:19 of 5-on-4 time once Roman Polak exited the box. This is very general since it doesn’t include the Blues’ penalty killers, but I like assuming opponents are equal and working around general expectations.

So, really, the Canucks’ had a .330 chance of scoring on the 5-on-3. If you include the minutes before the Pietrangelo penalty, Vancouver had a .508 chance of scoring on the sequence as a whole. But before the 5-on-3, nobody would have expected it to be a real make-or-break moment for the Canucks. I found it interesting because my father and I both had significantly different predictions and significantly different reactions to the powerplay when it ended up being killed off. Predictably, my reaction was that the odds were against the Canucks, and not to blame them or their coaching or any other reason for the failure of that powerplay. It would be a lot of like blaming a wrist injury for failing to mount a successful defence of Kamchatka against an attacking army from Alaska that has many more plastic cannons and horses.

The moral here is…don’t think that any moment is “now or never” or the like. Chances are, the Canucks’ won’t score. Even if they’re a good offensive team, the weird thing about hockey is that in any given moment, with any given player on any given spot on the ice, there’s no spot where a player is expected to score more than 50% of the time. That’s why successful teams are the ones that create offensive zone time, extend offensive zone time, and in the end, generate more scoring chances than the other team.

Despite the hiccup against St. Louis, they’ve been firing at a rate of over a goal per two minutes played 5-on-3 this season. This is something that’s a historical anomaly since the Canucks, contrary to the rest of the National Hockey League, score more goals at 4-on-3 than 5-on-3:

  5-vs-4 Goals/2 MIN 4-vs3 Goals/2 MIN 5-vs-3 Goals/2 MIN
2009 0.196 1.348 1.024
2010 0.245 0.774 0.644
2011 0.310 0.532 0.281
2012 0.231 0.698 0.500
2013* 0.198 n/a 1.091

*note the sample size this season. The Canucks have played just 1:50 of 5-on-3 time and have scored, and haven’t played a second of time 4-on-3.

What’s interesting about this next table is, contrary to the NHL average, the Canucks do much, much better at 4-on-3, as mentioned above, than 5-on-3. At one point my father said “the Canucks should have pulled Luongo on the 5-on-3” but I disagreed, noting that the Canucks’ powerplay works better with more space. I have no idea why this is:

  5-vs-4 Goals/2 MIN 4-vs3 Goals/2 MIN 5-vs-3 Goals/2 MIN
Canucks 2008-2012 0.243 0.785 0.664
Canucks 2011-2012 0.271 0.604 0.443
NHL 2008-2012 0.206 0.422 0.690
NHL 2011-2012 0.199 0.416 0.739

Data collected from tallying up numbers found at NHL.com.


  • JCDavies

    have you compared these numbers to teams that historically have had good powerplays? maybe this is symptomatic of having, say, 4 elite forwards or three elite forwards and an elite defenceman and the 5th guy is literally better off playing in the neutral zone?

    like if you put an aaron rome on the powerplay it wouldnt be shocking to say his best spot is somewhere between the defensive hash marks and the red line

  • JCDavies

    “the data is quite clear that the length of a powerplay determines quite a bit. 5-on-3s are successful at a 28% rate, but that’s because the average 5-on-3 is :46 long”

    This is really interesting.

    Its not something I’ve thought a lot about but it must be subject to the same inherent problems I would imagine all PP/PK stats are subject to. Specifically, the fact that good powerplays score more often and therefore stop the clock where as poor playerplays don’t score as much and don’t stop the clock as often – resulting in time data that is skewed towards the poorer powerplay stats. And, also, the significantly smaller sample size compared to 5v4 powerplays.

    I wonder if a weighted system based on 5v3 success rates might reveal any trends. Or, perhaps, dividing the teams into 3 groups based on success rates (good, poor, and average) and looking for differences between the groups.