(SPOILER: The answers are “mostly no” and “probably.”)
One of the biggest catalysts for advanced statistical analysis in baseball was the availability of play-by-play data, and the same is certainly true in hockey. In recent years, the NHL has published their own version of the play-by-play (see, for example, last night’s Winnipeg-Montreal game) and I finally have my hands on all these play-by-play files from last year’s regular season.
I am not a hockey analyst by trade, so I asked some others what they wanted to learn from the play-by-play. One suggestion:
“See if there are any players with the ability to create rebounds (based on a percentage of shots) and which goalies give up the most rebounds.”
That study follows.
Score effects affect the flow of play in hockey, as do man-advantage situations, so we’re only looking at 5-on-5, score-tied situations here from 2010-11. Unless otherwise noted, all statistics are from that game situation only.
Overall, we found 57,477 shots on goal. There were 4,503 goals, so we had 52,974 shots on goal that did not go in. Of those shots, 2,021 (3.8%) came within three seconds of a previous shot — the time interval we’re using to define a rebound, per Gabe Desjardins’ earlier work. (From now on, “shots” means “shots on goal that were not goals” — essentially, saves).
There were 622 shooters with at least 25 shots, including players on multiple teams — like Kevin Shattenkirk, whose time in St. Louis saw him generate 5 rebounds on 36 shots (13.9%), or my man Ian White* who had 5 on 37 (13.5%) with San Jose.
* For those familiar with the Jeff Skinner Rankings, I did a retroactive version of them for the 2002 draft. Ian White was very high in the rankings that year; I was certainly higher on him than NHL teams were, based on his draft position (191st). Despite being drafted near the end of the sixth round, likely due to his smaller size and “lack of physicality” in junior, White has gone on to outscore 51 of the 60 CHL skaters drafted ahead of him, many of whom of course never made the NHL. White has now caught on with Detroit, is playing top-line defence minutes, and was recently near the top of the Corsi rankings. Shorter version: Ian White is my new favourite player.
But we need to regress these observed rebound rates an appropriate amount to determine “true talent” levels here. Lifting yet more methodology from others, it appears you need to add 253 shots of league-average rebounding rates to a player’s results to estimate the skill. So a player with 100 shots needs to be regressed 71% of the way towards the mean.
And when we do that, we get this list:
Skaters, best at generating rebounds
6.4% S.J #8 PAVELSKI
5.8% S.J #44 VLASIC
5.8% NYI #91 TAVARES
5.7% PHI #17 CARTER
5.7% CBJ #71 CLARK
5.6% T.B #12 GAGNE
5.4% WSH #22 KNUBLE
5.4% CGY #20 GLENCROSS
5.4% BUF #36 KALETA
5.4% S.J #39 COUTURE
(The names come from the play-by-play description and to save time, I left them as-is, with team and number.)
So what does it mean to be Joe Pavelski on this list? His 6.4% vs. the league average of 3.8% means he’s generating 2.6 more rebound shots per every 100 of his shots. (He may pick up some of those rebounds himself.) If we assume a rebound shot is 40% likely to go in — see Gabe’s work again — then that’s about one more goal generated by Pavelski per 100 shots. He had 176 shots last year, meaning we estimate that he produced nearly two more goals due to his rebound generations alone.
How about the worst rebound generators?
Skaters, worst at generating rebounds
2.8% DET #40 ZETTERBERG
2.8% T.B #16 PURCELL
2.7% TOR #16 MACARTHUR
2.7% L.A #28 STOLL
2.7% NSH #4 FRANSON
2.6% VAN #26 SAMUELSSON
2.6% TOR #3 PHANEUF
2.6% T.B #4 LECAVALIER
2.6% PHX #17 VRBATA
2.5% MIN #22 CLUTTERBUCK
Fortuitously, we have Lightning teammates on each list, so we can compare the difference between Vincent Lecavalier and Simon Gagné without worrying about any inter-team factors. 5.6 minus 2.6 is 3.0 shots per 100, or 1.2 goals per 100 shots. Lecavalier took 113 shots and Gagné took 101, so it’s about a goal overall more for Gagné, to make up somewhat for his eight-goal deficit vs. Vinny. One goal over the course of a season isn’t going to turn a third-liner into a star, but it’s something.
The difference between the best (Pavelski) and worst (Cal Clutterbuck) is 1.6 goals per 100 shots, or .016 subtracted from the opposing goalie’s save percentage. That does make a difference, not that anyone’s about to give Pavelski’s minutes to Clutterbuck.
We’ll do the same thing as above for goalies. 56 goaltenders made at least 300 saves, and the spread in observed “rebounds allowed” leads us to determine that you need to add 1,772 saves at league-average rates to estimate skill. A goalie facing 1,200 5-on-5, score-tied shots therefore needs to be regressed somewhat less than the skaters with 100 shots — only about 40% of the way towards the average.
When we add those 1,772 shots to each goaltender’s actual rebound numbers, we get this list of the best and worst:
Goalies, best at suppressing rebounds
3.2% JONATHAN QUICK (L.A)
3.3% JAMES REIMER (TOR)
3.4% STEVE MASON (CBJ)
3.4% JIMMY HOWARD (DET)
3.5% ILYA BRYZGALOV (PHX)
3.6% ROBERTO LUONGO (VAN)
Goalies, worst at suppressing rebounds
4.1% MARTY TURCO
4.1% RICK DIPIETRO
4.2% CAM WARD
4.2% NIKLAS BACKSTROM
4.3% SERGEI BOBROVSKY
4.7% JONAS HILLER
So Jonathan Quick or James Reimer allows 0.5 fewer rebounds per 100 shots, or about 0.2 goals per 100 shots, or +.002 in save percentage. That’s maybe three goals for a true full-time goalie, call it two goals for someone who doesn’t play every game.
However, it’s not like being really good at preventing rebounds makes you a much better goaltender, or vice versa. For all goalies, the correlation between 5-on-5 save percentage and rebounds allowed is pretty much zero (0.02). And the “trailers” listed above averaged a .918 (5-on-5) last year to the leaders’ .916, so there also isn’t evidence that the big rebound allowers are also the worst goalies.
Team-wide rebound stats
You are no doubt also wondering about the team-wide stats.
While the difference between the top and bottom is huge (3.2 shots per 100, or 22 goals over the course of a season), this may be due to arena bias; different places count shots differently. We also see teams like Colorado and Florida at the top, and Chicago and Detroit and Anaheim at the bottom, and there’s little correlation (0.10) between rebound rate and a measure like score-tied Fenwick (5-on-5 shot differential ignoring blocked shots).
The best team, the Sharks, generate 2.1 more rebounds per 100 shots compared to the average, or 14 goals per season.
And on the other side of the ice:
Again, arena bias. A typical workaround is to look only at road stats for each team, but we’ll put that under “future research.”
Still, here, the difference between the best and worst is about 0.7 goals per 100 shots, or .007 in save percentage. A team’s goalie rebound rate correlates with 5-on-5 save percentage not at all (0.009), even less than the shooters (0.010) above.
The best team, the Kings, saved themselves about five goals vs. the average team over the course of a full season.
Basically, while we aren’t getting a list of the best and worst shooters and goaltenders here, we are finding that generating rebounds is worth about two goals for the best individual skaters, and about two goals for the best goalies. Yet that skill is, likely, mostly unrelated to the “bigger” things in the game, like controlling shots or simply being a better player.
If you took an average team and made them the best at both generating rebounds and avoiding giving them up, while keeping everything else the same, they would, by the above analysis, score 14 more goals and allow 5 fewer goals — so they’d be 19 goals better, which we’ll call about three wins in a season. That only really matters if you can find players who are good enough otherwise to justify giving them all that playing time, of course. Tim Thomas, for example, makes enough saves to overcome his merely average rebound suppression, and vice versa for Steve Mason.
Still, three wins is a huge effect, and when combined with the small results for individual skaters, probably suggests that rebound generation is a team skill on the offensive level, and more of an individual goalie skill on the defensive level. In other words, individual skaters cannot generate many more rebounds, necessarily, but a team-wide offensive strategy or system can. That’s probably what’s going on with San Jose (if it’s not arena bias), who had five players in the top 20. Goaltenders do seem able to control rebounds more than shooters can produce them, which only seems natural.
In short, there definitely seems to be an individual goaltender skill in avoiding rebounds, but whether that skill matters is up for debate.