Last week, the Panthers and Capitals set a record by taking a shootout 20 rounds, eclipsing the previous record (15) held by a November 2005 game between the Caps and Rangers. It was also the first time since the shootout was implemented in 2005/2006 that skaters were allowed a second attempt, which is only permissible after all 18 skaters on a team have attempted a shot. But beyond the flat unlikelihood of a 20-round shootout, there was another perhaps more interesting quirk in the contest.
After the game went to sudden-death in the fourth round, there were five instances in which the Panthers, who shot second in each round, surrendered a goal in the top half of the round and had to score to keep the game alive. They scored in all five of these instances. On the other hand, Florida had thirteen attempts in which a goal would have ended the game. They failed to score on every one of these attempts until the last one in round twenty.
How unusual then was it that the Panthers went 5-for-5 on must-score attempts, but only 1-for-13 on possible game-winners? More generally, are there differences in the goal scoring rates between when a team must score to avoid a loss, when a team can win with a goal, and when an attempt does not guarantee a victory or defeat? Sentiment seems to lean toward teams tending to score more frequently when they absolutely must, but there's a way to answer this a little more concretely.
I collected data about every NHL shootout attempt from the 2007-08 through 2013-14 seasons. This gave me just over 1100 shootouts to analyze. In these 1100 games, there were 1194 attempts in which the team could win with a goal, 935 attempts in which a team had to score to avoid defeat, and 5548 attempts that could not end the game either way.
The graph above shows the percentage of times a goal was scored in each of the three shootout circumstances. The thick lines show the 90 percent confidence intervals; the thin lines denote 95 percent. The red and green bars are slightly lower than the yellow one, suggesting that goal-scoring rates may diminish when there is more pressure on the shooter.
A simple t-test, which evaluates whether these means are statistically distinguishable from each other, indicates that the goal probability in low pressure situations (yellow) is statistically significantly higher (p<.05) than the goal probability when a team must score to avoid a loss (red). However, there are no significant differences between the goal probability when a goal wins the game and either of the other two scenarios.
This result should not be especially surprising. A team that faces a must-score situation got there precisely because they were scored fewer goals than their opponent up to that point. Therefore it makes some sense that the goal scoring rate would be lower in these situations.
The difference in these rates may be further explained by who the home team is. The figure above divides out the goal scoring probabilities by both the game situation and whether the shooter is at home or on the road. If you look just at the bars for the home team, you will see their probability of scoring a goal does not seem to be affected by the game situation.
The differences in goal scoring rates across game situations appear to be mostly driven by the away team. The away team's success rate is significantly better (p<.05) in the lower pressure situation (yellow) than in either of the higher pressure ones (green and red). One possible explanation for this is that the home crowd tends to be louder and more hostile when the away shooter is taking a game-deciding attempt.
What do these results tell us about the likelihood of seeing another 20 round shootout sometime soon? We know, empirically, that 31.4 percent of shootouts are tied after the first three rounds. We can also use basic principles of probability to calculate that once the game goes to sudden death after the third round, there is a 55.7 percent chance that the game advances to another round.
Based on these two statistics, the probability that a shootout lasts 20 or more rounds is about 0.0027 percent. So enjoy the video of Tuesday's shootout because we're not likely to see another one that long for quite some time.
Stephen Pettigrew is a PhD candidate at Harvard University, where he studies political science and American politics. He also has a master's degree in statistics from Harvard. In his spare time, he writes about sports analytics, particularly in hockey and football.