He Puzzled And He Puzzed

How many times have you been watching your team carry a lead, only to have the commentators announce that the opposing team has scored most of their goals in a later period? Listen, I get it; the talking heads on TV have to maintain a certain level of suspense. They can’t very well say "It’s looking pretty unlikely that this game takes a turn for the interesting." And although I love mythbusting when it comes to hockey, I will concede that sometimes narratives are more fun to build than reasonable arguments.

But, as any of you who are familiar with my work will already know full well, I’m never content to accept in-game narratives without some kind of verification. 

So what about teams that score more goals in the third period than any other? Are they successful? What about those teams that score most often early in the first? I took a look at 6 seasons’ worth of data to check.

My own hypothesis when it comes to this is based on Gabriel Desjardins’ posts on score effects. If a team is trailing, they are more likely to shoot more in an attempt to catch up on the scoreboard. It would seemingly follow then, since shots are the biggest statistical drivers of goals, that teams that score most often in the third are the teams trying to catch up. Basically, I expect that poorer teams score later. Right?

This first set of graphs show the relationship between goal differential in a given period and where the team wound up in the standings at the end of a season:




It seems pretty obvious that the more a team is out-scoring their opposition in any given period, the more likely it is they win games, so the fact that each graph shows a strong correlation isn’t surprising. But what is more intriguing – and what supports my hypothesis – is that the connection is strongest in the first period, and drops as the game progresses. Score often, sure, but scoring early and often, is best, right? Well, here comes the wrinkle.

I divided every team season for the last six years into three groups: those that had their best goal differential in the 1st, those that had their best goal differential in the 2nd, and those who did in the 3rd, and looked at the average number of points they achieved in the standings. Here’s what I found: 

So what the heck is happening here? It seems that the second period is (marginally) favored by winning teams, and that a whole lot of non-playoff teams make their best period the first one. Why doesn’t it adhere to the score effects we’ve seen on shots? Of course, this doesn’t mean that good teams get out-scored in the first, it just means they wait to fill their opponents’ nets with pucks. But why the wait?

For now, I wonder if Home and Away games make a difference, or if Lady Luck might also have a lot to do with this. But one other idea I’ve had is that the longer a team has to play with a lead, the longer the trailing team will have spent sending an increased volume of shots at the net, and therefore, will find the time to win if they go down very early in the game.

In any event, with the numbers put thusly, it appears that there are simply teams that have been able to win by coming on strong towards the end of the game, and that they’ve been able to do it more consistently than others throughout a season.

The fact that many teams have repeated their pattern of scoring more goals in a given period lends a modicum of feasibility to this idea. Detroit, for instance, scored most in the 3rd period 5 out of 6 seasons. Of course, models of terribleness the Atlanta Thrashers did it 4 out of 6 times as well.

My gut feeling is that there’s nothing to this strange stat, but I can’t totally discount it, either. Any thoughts?

  • ChinookArchYYC

    Interesting read, especially the results in the 2nd period, over the last 6 seasons.

    Could it be that bad teams That play with very little confidence give up? How many times have you watched a struggle team play tight in the early part of a game (I.e. the 1st period), keeping things close until they fall behind? In a league with as much parity as the NHL, this ‘giving up’ effect divides bad teams from mediocre teams and mediocre teams from great teams.

  • Marda Miller

    clutch/60 increases in the 2nd and 3rd, obviously.


    I agree with your explanation of 3rd period scoring, but I just can’t wrap my head around how it fits into the second period.

    what I would like to know is how many first periods end in ties and if it’s disproportionally larger then that stat in at the end of the second.

  • Alex Hemsky

    Something else that strikes me as interesting: The later the period the more outliers in scoring differential there are.

    Per period the number of data points outside of +/- 40 goal differential are:

    1st 2
    2nd 5
    3rd 9

    I assume empty net goals have not been excluded. Would that account for some of the increase in the 3rd period?

  • Marda Miller

    Couldn’t the explanation be that better teams are more able to expand or contract playing to the score effects, so as to make overcoming randomness earlier in the game more likely?

    Imagine two teams. A is a good team, B is a bad team. A is more likely to have more goals in periods 1 and 2. But also, they are more able to suppress shots in period 3 when ahead and less likely to be suppressed in period 3 when behind. As they are leading most often, they still have a positive score differential in the 3rd because the almost never get scored on. When they have had bad luck in period 1 they are able to ratchet it up and still score in the 2nd or 3rd. The differential in the 3rd is exacerbated for good teams.

    For B, on the other hand, if they fluke into a lead in period 1 they sit back in periods 2 and 3, but because they suck they often get shelled and scored on. When they are trailing, as they often are, their superior opponents are able to completely shut off any meaningful possession.

    Good teams do even better than poor teams in period 3 as the score effects exacerbate their differences.

  • Marda Miller

    This is a reach, but perhaps teams require a few minutes to get fully engaged with each other. Thus the first period would act as a partial grace period in some games, with the scoreless firsts boosting the goal differentials of bad teams.

  • Marda Miller

    I wonder if the difference in optimal period goal differential might be a measure of intra-game coaching effectiveness. That is, better coaches, can see the strategy the other team is using and how poorly their own team is playing against it and then suggest changes to the team playing style to counter it during the intermissions