An analysis of the timing of goals in Hill-Murray hockey games, 2004-2015

By Ryan Shanley

 

Hill-Murray is a high school in Maplewood, Minnesota, whose hockey team is usually among the top ten out of about 150 schools in Minnesota. Its most well-known recent alumnus may be Jake Guentzel, who debuted in the National Hockey League with Pittsburgh in 2016.

This is a statistical exploration of the timing of goals. Coaches often say that it’s especially bad to give up a goal in the first or last minute of a period. Regardless of whether that is true, can we at least determine whether goals are more or less likely at certain times during a game? This is about broad long-term trends; there is so much variation in individual games that one could not use this to predict single game outcomes.

I have a database of game records that is complete from the 2000 season onwards. However, period length increased from 15 to 17 minutes in 2004, so I’m going to exclude games prior to the year. Through the middle of the 2015 season, that gives us 342 games, and 2476 total goals by both teams not counting overtime. I’m going to ignore overtime for the most part; there aren’t a lot of overtime games - not many goals to analyze anyway - but I’ll show what’s there at the end.

For each goal, the time in minutes and seconds since the start of the period is recorded. There may be recorder error in some cases, such as off by 10 seconds or 1 minute. I estimate that this occurs in < 10% of cases, but have no way of knowing which ones and doubt the errors are systematically biased in one direction or another.

As a sanity check, let’s look at the distribution of seconds, which should be fairly uniform. (This is irrespective of whichever minute the goal occurred - for this exercise, 59 seconds means the goal could have been scored at 0:59, 1:59, 2:59, etc.)

 

 

Hmm, it’s a bit puzzling that the highest value occurs at 0 seconds, and there are some spikes at other round numbers, like 10, 15, 20, and 30.

There is a mercy rule in effect whenever a team is leading by six or more goals in the third period. The game is played on “running time” in which the game clock does not stop when the puck is out-of-play. What I think could be happening is that when the game is played with running time, the scorer rounds the goal time to the nearest 5 or 10 seconds, and so we see an excess of goals at x:00, x:10, etc.

And 19 seconds seems to be a goal scoring void for some reason.

Now on to more interesting topics. Here are goals plotted by period and minute - i.e. the 0 minute is 0:00 to 0:59. You could call it the first minute, but that’s the convention I’m using.

 

 

The mean is 7.2 goals per game. In a 51 minute game, that equates to about one goal every 7 minutes. Multiplied by 342 games, we get an average of 49 total goals per distinct minute in this sample.

There were 772 goals (31%) in the first period, 897 (36%) in the second, and 807 (33%) in the third. The highest scoring minute was the final minute of the third, which should be a tip-off that it’s inflated by empty net goals. There were 38 empty net goals, 33 in the last minute, so without those, scoring at the end of the third is pretty much in line with the rest of the period. In fact, if we take out the empty net goals, the third period is almost identical to the first in total scoring. The lowest scoring minute is the first minute of the third period with only 30 goals, almost 40% lower than average.

Total Goals
First 8 minutes Last 8 minutes
Period 1 352 363
Period 2 383 466
Period 3 354 360

(excluding the “middle” minute and empty net goals)

Second period scoring is 8% higher in the first 8 minutes compared to the first and third periods, which could be a chance difference. But it’s 29% higher in the last 8 minutes.

It might be interesting to distinguish “blowout” games from “competitive” ones. I defined a blowout as Hill-Murray scoring 7 or more goals (there were no blowouts in the other direction), which includes 95 games (28% of the total). Not surprisingly, goals per game were higher in those games: 10.5 versus 6.0 in the 247 competitive games.

 

 

As expected, the empty net goal blip disappears in blowout games. We see a slight increase late in the second period by HM, and a moderate dropoff in the third - understandable since the game is probably in hand by this point and could be played on running time.

Goals per blowout game
Period 1 Period 2 Period 3 Total
Hill-Murray 3.28 3.53 2.76 9.57
Opponents 0.25 0.41 0.26 0.92

Now for the 247 competitive games…

 

 

Goals per competitive game
Period 1 Period 2 Period 3 Total
Hill-Murray 1.09 1.34 1.30 3.73
Opponents 0.68 0.77 0.81 2.26

 

When we take out empty net goals, Hill-Murray does appear to have a second period boost on the order of 20% relative to the rest of the game, regardless of strength of opponent, and the difference is concentrated in the latter half of the second period. But I don’t want to make too big a deal of it, since the main theme is that goal scoring is fairly uniform throughout the game.

(There is a plausible reason why scoring would be highest in the second period - it is the only period in which teams defend the net farthest from their team bench. Therefore it is harder for tired players to change during play, and may create more scoring opportunities. This hypothesis could be tested for other teams and leagues.)

About 20% of all goals are scored on the power play. I’m not showing these separately because their distribution roughly follows total goals. However, there are about 25% fewer power play goals scored in the first period relative to the other two. I’m guessing that is mostly because there are no carryover penalties at the start of the game. Maybe there are fewer penalties called over the whole period, but I don’t have that data.

(I did a back-of-the-envelope estimation of what power plays do to the goal scoring rate, and concluded that it appoximately doubles the scoring rate for the team with the advantage - maybe a little more, like 2.5x, while driving the other team’s rate close to zero. That results in a net increase in combined goals and obviously a drastic shift in the likelihood of which team scores.)

Finally, here are the first overtime goals. (Overtime periods are sudden-death with an 8 minute maximum time.) There were 22 goals in 35 games (the 13 games in which no goals were scored included 8 ties and 5 playoff games decided in multiple overtimes). We should expect a decreasing trend in the number of goals in each minute, since a game won’t reach a later minute if a goal is scored beforehand.

There were 14 goals in the first four minutes versus 8 in the last four, check. The goal scoring rate was about 10% higher in overtime relative to regulation play in competitive games, more similar than different.