The other day I wrote about Rielly and his on-ice save percentage. A brief twitter discussion with Tyler Dellow occurred which ended with Dellow asking me to explain how Rielly went from -0.9 Sv% Rel in 2015-16 to -2.8 Sv% Rel in 2016-17 while playing similar shut down roles in both seasons.
He was -.009 relative to team in 15-16 playing the same “shut down” workload as most of last year. You think he got lucky?
— dellowhockey (@dellowhockey) June 20, 2017
Let me go through a couple possible explanations.
Change in Defense Partners
In 2015-16 Rielly’s main defense partner was Matt Hunwick while in 2016-17 it was mostly Nikita Zaitsev. We only have one season of data for Zaitsev so it is difficult to get a sense of what his impact on save percentage may be but we have several years of data for Hunwick.
In 5 of the past 6 seasons Hunwick has had a positive Sv% Rel with the only negative season being the one he played with Rielly. I think this suggests that there is a very good reason to believe that Hunwick has a positive impact on save percentage and could very well have pulled Rielly’s on-ice save percentage up last year.
|Rielly and Hunwick||91.47|
|Rielly without Hunwick||91.22|
|Hunwick without Rielly||93.45|
Looking at their WOWY’s it seems Rielly pulled down Hunwick’s save percentage more than Hunwick pulled up Rielly’s however Rielly was still a little better playing with Hunwick.
The Leafs tanked in 2015-16 and traded away several notable players, including Dion Phaneuf and Roman Polak. I have written about both of these guys in the past and their ability to boost on-ice save percentage. Phaneuf has a history of being able to boost on-ice save percentage and Polak has thrived at doing so under Mike Babcock. Both of these guys were traded in February, 2016. Phaneuf was traded February 9th and Polak on February 22nd. Using February 22nd as a cutoff we can see how Rielly performed before and after these trades.
|Rielly On||Rielly Off||Sv% Rel|
|Prior to Feb 22||91.06||93.14||-2.08|
|After Feb 22||92||90||+2.00|
Not surprisingly when the Leafs traded two good save percentage players (both of whom Rielly rarely got to play with) Rielly’s off-ice save percentage tanked resulting in a big boost to his Sv% Rel. When the Leafs were actually fielding a real NHL defense Rielly’s Sv% Rel was -2.08, which is much closer to where he was this past season.
Phaneuf and Polak were replaced by Corrado and Carrick, neither of whom appear to be particularly good at boosting on-ice save percentage.
Change in Score Situation Rates
In my previous article I showed that in 2016-17 Rielly had the worst Sv% Rel in 5v5 leading situations. The same is true for the 2015-16 season. It should be noted that the Leafs were terrible in 2015-16 and led games far less frequently than this past season when the Leafs made the playoffs. This can actually have a fairly significant impact on Sv% Rel.
|Sv% Rel||Sv% Rel||TOI||TOI||% of 5v5 TOI||% of 5v5 TOI|
Rielly’s Sv% Rel was better in each of the leading, tied and trailing categories in 2015-16 than 2016-17 For 5v5 leading and 5v5 tied the difference was much smaller (0.7 and 1.0 respectively) than difference in 5v5 Sv% Rel (1.9). The difference in 5v5 trailing situations was slightly larger (2.1).
I do find it really interesting that where Rielly was bad in 2015-16 he continued to be bad in 2016-17 and where he performed the best in 2015-16 he performed the best in 2016-17 as well. Luck and randomness don’t normally work like that.
Also take note of the differences in the percentage of 5v5 ice time that Rielly was on the ice when leading, tied and trailing. In 2015-16 just 21.8% of his 5v5 ice time was when the team was leading. That jumped to 40.5% in 2016-17. His ice time when trailing dropped from 37.8% in 2015-16 to 26.3% in 2016-17.
These differences in ice time do make a difference. If we took Rielly’s 2015-16 Sv% Rel stats by score situation and assumed he played 2016-17 score situation percentages his Sv% Rel would be -1.60 which is a fair bit worse than his actual -0.9 Sv% Rel. (The math goes like this: -3.9*0.405 + -1.0*0.332 + 1.2*0.263 = -1.60)
The change in score situation accounts for about 37% of the difference between his 2015-16 and 2016-17 Sv% Rel.
In hockey analytics we talk a lot about regressing to the mean, but rarely do we really explore why players regress. Too often we chalk it up to luck and randomness but that is not always the case. Luck and randomness do exist, but they shouldn’t be used as a catch all excuse for year to year changes in observations. I just explored 3 viable reasons why Rielly’s Sv% Rel got so much worse this year. The combination of all three could very well explain the change Rielly’s on-ice save percentage.
So, when Tyler Dellow asks me to predict what Rielly’s on-ice save percentage will be next year I have a really hard time making a prediction. I simply don’t know how he will be used and whether he will change defense partners. Will the Leafs be able to acquire a defenseman capable of handling the tough minutes or will Rielly be depended on to do that again? Will Hunwick and/or Polak be back? Will the Leafs be even better resulting in even more 5v5 leading ice time which is when Rielly at his worst? Without knowing the answers to these questions it is really difficult to predict what Rielly’s Sv%Rel will be next year.
What I am confident in saying is that Rielly should not be used as a shutdown defenseman. Let him lead the second pairing in a more offensive role and sheltered from the real tough minutes. If I am being perfectly honest, I’d even consider trading him if it was necessary to acquire a true top pairing defenseman.