Can goalies control the number of rebounds they allow?

Rob Pettapiece
July 15 2013 01:42PM

 

Pekka Rinne
- pic via Mel

 

A while ago, I looked at whether skaters (or goalies) had any ability to generate (or avoid allowing) rebounds. With only one year of data, I concluded that "there definitely seems to be an individual goaltender skill in avoiding rebounds, but whether that skill matters is up for debate." This is an updated and improved version of that study, focusing just on goaltenders.

Methodology

We will define a rebound as any shot taken at 5v5 within three seconds of a previous shot, in a continuous action situation. These are known to be dangerous; see here and here for examples of the very low save percentages on rebound shots. That second link tells us that rebounds account for 11% of all shots taken within 60 seconds of the previous shot, but 34% of the goals. So — and this is not exactly earth-shattering — a goaltender who has the ability to cover up rebounds, or control them enough that a teammate can clear the zone, can significantly increase his value to his team.

A goalie's rebound rate is then the number of rebound shots he allows, divided by the number of saves he makes, such as James Reimer's 2.3% rate on 841 saves in 2010-11, one of the best seasons on record for rebound suppression, or James Reimer's 6.0% on 761 saves in 2013, one of the worst seasons on record for rebound suppression. (That should give you an idea of the variability in this statistic.)

Results

At this point we'll partially obey the law of headlines and say: no, outside of a few cases, goalies generally do not have an ability to control rebounds.

It should be noted that what we really mean here is: given the play-by-play data available to us, goalies generally do not have an ability to control rebounds. If you track your own statistics during a game, and then compare them to the NHL's official record, more often than not you'll find discrepancies. Shots are recorded at the wrong time or credited to the wrong player, or even (rather famously) the faceoff dots are all over the place. All reported results here (and anywhere that the play-by-play is used as heavily) are subject to that disclaimer.

I could tell you that Reimer had a 2.3% rebound rate one year, then 4.0% the next, then 6.0% in the next, but is that really telling you he got worse? (League average is around 5.2%.) Pekka Rinne was at 3.0% in 2011-12, but 4.7% the year before and 3.6% the year after. How much of those changes are due to the goaltenders themselves?

One way to look at that is to group goalies by their performance in one year, then look at how they all did the next year. We have 79 goalies who saved at least 300 shots in consecutive years, and here's what their year-to-year rebound rate looks like:

Group Year 1 Year 2
1 3.8% 4.8%
2 4.7% 4.8%
3 5.1% 5.4%
4 5.6% 5.6%
5 6.4% 5.6%

The regression is severe for the top group: more than 70% of the way towards the average for Group 1, which is part of the reason that I don't think goalies can really suppress rebounds that much. But the goalies who were not good in one year don't regress as much, possibly because they are truly that bad, or possibly because the really bad/unlucky ones didn't get a chance to put up more mediocre numbers in year 2 and thus increase the year-to-year improvement of the group.

The "Pekka Rinne Rankings"?

Given the variability in rebound rate, we need to add about 570 league-average saves to everyone's totals to get a "true" estimate of their rebound-suppression skill. When we add up the last three years of data, and regress appropriately, we get the following list, sorted by "Reb GAA" — goals above average with respect to rebound rate only.

Name Shots "True" Reb% Reb GAA
Pekka Rinne 4170 3.9% 19.8
James Reimer 2492 4.3% 8.7
Michal Neuvirth 1995 4.2% 7.7
Ilya Bryzgalov 3651 4.6% 7.7
Ryan Miller 3857 4.7% 6.8
Tim Thomas 2754 4.6% 6.7
Jonathan Quick 3372 4.7% 6.5
Tuukka Rask 1987 4.4% 5.9
Mike Smith 2832 4.6% 5.8
Roberto Luongo 3036 4.7% 5.6

These goalies were all able to add a win or more through rebound suppression, compared to an average goalie, and they are mostly good goaltenders overall. (The 11th and 12th on this list are Craig Anderson and Carey Price.) Of course, if you let up fewer high-percentage rebounds, you'd have a better save percentage, too.

Rinne blows away the field here, which might be appropriate given what people have noticed about his ability with the glove. Shea Weber describes Rinne's skills so well at that link, I'll just yield the floor to him here:

He eliminates a lot of second chances himself [by] smothering pucks with his glove and he’s so good at catching and scooping pucks off the ice. I’d never seen it before until I started paying with him, first in Milwaukee. He would scoop those pucks right off the ice if you were shooting for his far side pad. It’s an unbelievable skill, because without a doubt that’s a rebound on any other goalie that could [possibly] come out for a second chance or another shot.

And there is some evidence that Weber isn't just speaking well of a teammate here.

Keep in mind, though, that we had to combine three years' worth of data to find this effect — or, rather, 2.6 years. The amount we have to add to a goalie's raw rebound rate at 5v5, 570 saves, is equivalent to about 27 games. So even the goalie who played the most this past year, Ondrej Pavelec, is regressed nearly half of the way to the mean, from his 6.6% to 6.0%. It's hard to identify this skill during the season, if it is even indeed a skill.

Looking ahead

For the goalies who played in consecutive years (2010-11 and 2011-12, or 2011-12 and 2013), the best way to predict their Year 2 rebound rate is simply to guess that they'll be league-average. But let's add to that sample, and see if we can find any predictive value to the regressed rebound rate.

Pekka Rinne's regressed rate was 4.3% this past year. We don't yet know what he'll do next year, but let's pair him with Braden Holtby, who had a similar 5v5 SV% in similar playing time but a 5.5% rebound rate.

We'll repeat that for everyone in the top 5:

Pekka Rinne (Braden Holtby)
Carey Price (Kari Lehtonen)
Mike Smith (James Reimer)
Marc-Andre Fleury (Viktor Fasth)
Craig Anderson (Sergei Bobrovsky)

A year from now, we can check the rebound rates of each group. If there's no difference, then it would be hard for us to say there's much value in one year's worth of rebound statistics for goalies.

We can also take that three-year top 10 list from above, and see how close those players are to league average next year. My guess is that the larger sample means they'd regress less, assuming no significant aging effects, but that is just a guess.

Conclusion

Aside from Rinne, who could very well be an outlier here (and we have reason to believe he could be legitimately great at this), there isn't that much a goalie can do in a given year to suggest he has the ability to save more than a few goals. This may be an artifact of the data quality we have to work with. Or it could simply be too difficult for most NHL goalies to control what happens to a very-high-speed projectile after it hits them in the chest.

Aa1ba7cac490a59c5fbd224ac4c13047
Rob Pettapiece is a sports analyst based out of southern Ontario. He edits The CIS Blog and provides statistical consulting services to front offices and journalists in various sports. He can be reached at rob.pettapiece at gmail dot com or @RobPettapiece.
Avatar
#1 Eric T.
July 15 2013, 02:52PM
Trash it!
0
trashes
+1
0
props

I've done some of this analysis and came to a slightly different conclusion. So I'll ask...

Setting the shots saved threshold as low as 300 shots still leaves you with a year-over-year persistence of 30% or more. That doesn't seem particularly low to me at all -- I doubt the repeatability of save% is all that much higher, and we certainly wouldn't say "no, outside of a few cases, goalies generally do not have an ability to control [save percentage]".

So why do you draw that conclusion about rebound%?

I agree that in any given year there's obviously significant variance, but I wonder if you're underselling things here.

Avatar
#2 Eric T.
July 15 2013, 03:05PM
Trash it!
0
trashes
+1
0
props

@Eric T.

I guess the answer is (in part, at least) that you don't see 30% persistence at the individual level; you see 30% persistence for the group, and likely a lot more noise for the individuals.

What's the year-over-year (or better: odd/even) repeatability correlation for the individual players?

Avatar
#5 Hand Knit
July 15 2013, 04:58PM
Trash it!
0
trashes
+1
0
props

I seem to remember reading that Nashville has a reputation for possibly overcounting shots. Could that have an effect on Rinne's numbers?

Avatar
#7 BenS
July 16 2013, 04:00PM
Trash it!
0
trashes
+1
0
props

Your group data does not support your conclusion. Yes, there's regression to the mean, but did you run a correlation on those numbers? I did, and it's 0.70. That's a strong correlation.

Also, even with the year-to-year correlation of 0.29 that you mention elsewhere in the comments, that accounts for ~9% of the variance. That's something I would pay attention to as an NHL GM. It's a weak correlation, but it's worth looking at over multiple years.

And, I hesitate to draw many conclusions that concern last year's season. The varying amount of preparation and compressed schedule introduce an awful lot of addition variables (as well as decreasing sample size).

Avatar
#8 BenS
July 16 2013, 04:15PM
Trash it!
0
trashes
+1
0
props

Sorry, I mistyped a number in my spreadsheet. The actual correlation for your groups are even higher (0.87).

Avatar
#10 Matt C
July 17 2013, 02:00AM
Trash it!
0
trashes
+1
0
props

Out of curiosity, did you define a rebound as a shot on goal or as a Corsi event/shot attempt/whatever it's being called these days.

Avatar
#11 BenS
July 17 2013, 08:11AM
Trash it!
0
trashes
+1
0
props

I didn't know your sample size before, but with 80 goalies, that 0.29 correlation is significant at an alpha level of 0.05. So there's that.

The fact that this was a lockout season also still really makes me wonder if this season was representative. I know you said that it's worth watching going forward; however, you might be able to accomplish the same thing by going back to earlier seasons. It would be a good way to add data without waiting for next season assuming the data exists.

Avatar
#12 Simo
July 17 2013, 01:20PM
Trash it!
0
trashes
+1
0
props

All those years playing Finnish baseball in school paid off for Rinne :)

Comments are closed for this article.