It is a well-known, and logical, fact of hockey that teams play to the score. A team protecting a one-goal lead in the third period will try to shut down the play, keep opponents’ shots on the periphery and keep scoring chances on both sides to a minimum. A team playing from behind will be much more aggressive, which in the extreme involves hockey’s last-ditch play, pulling the goalie. Teams do, and should, play to the score. What I did not expect is that they also shoot to the score.
In my previous article, I mentioned that the two biggest factors that affected the odds of a shot going in were the distance of the shot and the situation under which it was taken. Hockey bloggers, particularly JLikens at Objective NHL, have pointed out that there is a correlation between team shooting percentage, team save percentage and the amount of time spent playing in the lead. Since correlation is not causation, I decided to find out if there was something there.
To figure this out, I calculated for every shot and every goal in the NHL last year whether the team that took it or scored it was leading, trailing or tied. I eliminated all pulled-goalie situations, as they tend to increase the shooting percentage of the leading team significantly. I then calculated the expected shooting percentage of all these shots, based on their distance and the situation.
Here are the results:
Results Shots/60 Goals/60 Expected SH% Observed SH%
Trailing 31.7 2.86 9.1% 9.0%
Even 30.0 2.56 8.9% 8.5%
Leading 27.0 2.66 9.2% 9.9%
These are large effects. Teams playing with the lead take 5 shots less per 60 minutes than teams playing from behind, a difference of over 15%. This is somewhat expected, as teams protecting a lead will play a more defensive style and try to keep shot quality against to a minimum. However, what is even more interesting are the shooting percentages: the least dangerous shots, objectively speaking, come when the game is tied, but they go in even less often than we would expect. The shots of leading teams, on the other hand, are much more dangerous: while they would have been expected to succeed at a 9.2% success rate, they in fact did so at 9.9%. With over 21,000 shots last year, that’s a difference of 130 goals.
What explains these effects? It’s not power-play situations: the distribution of power-play and shorthanded time is random according to the game state: teams playing from behind don’t get take more or less penalties than teams playing with the lead. A more likely culprit is the type of game situation that arises. Let's say Boston is defending a 1-goal lead against Montreal, and Montreal is putting up a lot of pressure. The Canadiens are trying to keep the puck in the offensive zone, while taking quite a number of shots. The Bruins eventually get to the puck and create a 2-on-1 situation. The eventual shot may only be taken from 30 feet, but it’s a dangerous shot nonetheless.
These results are somewhat biased by the fact that better teams tend to get the lead, and better teams also have better shooting percentages. I therefore normalized the results assuming that every team played the same amount of time in all 3 situations and extrapolated their results. The final numbers are almost the same:
Norm Shots/60 Goals/60 Expected SH% Observed SH%
Trailing 31.9 2.89 9.1% 9.1%
Even 30.1 2.57 8.9% 8.6%
Leading 26.8 2.64 9.2% 9.8%
The shot differential gets even larger, as good teams become more represented among the trailers. The shooting percentage bias shrinks, but is still quite present. For those who are wondering, these effects were as prevalent the previous year as well:
Results Shots/60 Goals/60 Expected SH% Observed SH%
Trailing 30.4 2.64 9.1% 8.7%
Even 28.9 2.48 9.0% 8.6%
Leading 25.9 2.58 9.4% 10.0%
In 2007-08, even shots for trailing teams were less dangerous, which made the goal differential less severe: while leading teams were outscored by 0.2 goals per 60 minutes in 2008-09, they were only outscored by 0.06 goals in 2007-08. This is what happens when we have two effects which are cancelling each other out.
What does this mean for player and team evaluation? It depends on what metrics you are using. Shot-based metrics like Corsi numbers will undervalue players who are used to protect a lead, especially on teams with a lot of leads. Last year, that was especially true of Zdeno Chara and Nicklas Lidstrom, who both hardly need another boost to their value. However, at the individual level, the chance is not very large: since a player will take 5 shots for every 60 minutes in the lead, Chara, who played about 300 minutes more with the lead than from behind, would have his Corsi boosted by +25.
For goaltenders, it means goalies on bad teams will see their save percentage decline, as their teams take chances to get back in the game and leaves them wide open to dangerous shots. I have been particularly hard on Vesa Toskala in the past, but the more I dig into the numbers the more I realize how much the Leafs hung him out to dry last year. For Toskala, the effect would boost his save percentage by 0.0013, which would translate to 2 fewer goals allowed.
So the next time your team is protecting a lead in the final minutes, you can relax a little bit more. Those shots may look dangerous, but they’re probably not going in.
Tom Awad is an author of Hockey Prospectus.
You can contact Tom by clicking here or click here to see Tom's other articles.