To save us all from the trainwreck that is unfolding in this thread, I decided to not go to the gym and not make dinner tonight in order to update this for 2017. So, my wife is now all mad at you by proxy for the poor health decisions I've made, but alas, here it is.
Oh, and by way of explanation and marginal introduction, my name is Nik, I founded numberFire, which is a sports analytics platform now owned by FanDuel; we work with about ten teams in various capacities by mostly focus on the consumer space, fantasy sports and the like. I know you don't care, but I thought it was worth mentioning in regards to my bonafides in talking about this stuff.
---
1: Teams
The first thing we need to do is to look at the Penguins and profile them from a statistical point of view. The easy way to do it would be to use goals for and against; the harder way would be to use Corsi and various other possession/advanced stats. For the ease of collection on my end, I chose numberFire's proprietary nERD values, which is a composite statistic that roughly calculates how many goals plus or minus a team is relative to a perfectly average one. A score of +1 (i.e: should beat an "average" team by a goal every night on neutral ice) is considered extremely elite; the Pens were a +0.73 this year, finishing 2nd to the Capitals at +0.89.
Without knowing what changes the Pens will make vis-a-vis free agency, coaching, etc. we'll have to assume that +0.73 is the constant, then take the component stats that form it (offense, defense, power play, etc.), adjust it for era and strength of opponents, and find an array of similar teams.
Vancouver Canucks 2010 96.63%
Chicago Blackhawks 2014 96.57%
Philadelphia Flyers 2012 94.53%
Vancouver Canucks 2002 94.18%
Vancouver Canucks 2003 93.43%
San Jose Sharks 2002 93.11%
Dallas Stars 2008 92.29%
Ottawa Senators 2004 91.46%
Boston Bruins 2011 91.30%
Chicago Blackhawks 2011 90.84%
---
2: Players
The second thing I did was to form the neural network player array for Murray, using the updated stats (GAA, SV, VORP, nERD) from this year. While it's important to note that not all shots and scoring chances are made equal, unless STATS Inc. or SportRadar dramatically improves their investment in visual tracking, it is generally impossible to quantify those differences.
What we can (and will) do is the simple stuff: adjusting for era, adjusting for potentially disproportionate amounts of power plays for or against, that sort of thing. There is also a marginal adjustment made for teams who have an unusually high or low amount of goals scored; this is an inference I'm making in regards to their playing style, which will ultimately have an effect on the goals against.
The last thing I did was to adjust each potential player to what their projected performance would be playing on the Pens in 2017; that is to say, if the comparable is Patrick Roy 1987, we'd take Roy's historical statistical profile as of that year and apply it to Murray playing for a team statistically similar how the Penguins are anticipated to play in 2017. Whew. That's a mouthful.
Jaroslav Halak 2010 97.13%
Semyon Varlamov 2011 96.67%
Roberto Luongo 2004 95.43%
Miikka Kiprusoff 2006 94.75%
Marty Turco 2007 94.23%
Mike Smith 2012 92.33%
Niklas Backstrom 2009 91.99%
Carey Price 2011 91.11%
Jimmy Howard 2011 90.94%
Kari Lehtonen 2012 90.54%
(There's more than this; I'm cutting it off here for length.)
Three caveats here off the bat: these comparisons are purely quantitative and have nothing to do with personality/style/emotional comportment, whatever. You can put stock into those things if you choose; for this exercise, they're somewhat irrelevant.
Second, it's important to note that while Murray is in his second-year, some of the comparisons are for goalies later in their career. This is because we're comparing it on a seasonal level and not on the basis of similarity at that point of their career. I originally attempted to do it on this level, and realized that the sample size at that isn't large enough to do anything significant - Murray is that much of an outlier. That alone should tell you an awful lot.
Last, most of the goalies are somewhat recent. This is because I included VORP and nERD, two advanced derivative stats that are somewhat common and proprietary to my company, respectively. Both require high-fidelity raw data to calculate accurately, and that data generally wasn't available until recently.
---
3. Projections
From here, we find the union of the data sets; that is to say, find the closest comparable player on the closest comparable team at the closest comparable point in their career progression, weight it, and play it out from there with a running weighted average. I feel a little bit less comfortable with this given the amount of exogenous things that can wrong with huge impact (injuries to either Murray himself or to Letang/Crosby/etc, dramatic shifts in NHL rules around equipment or goal sizes) but this should probably give some initial guidance.
Sergei Bobrovsky 2013 94.45%
Tuukka Rask 2012 93.41%
Miikka Kiprusoff 2006 93.03%
Ryan Miller 2010 92.56%
Roberto Luongo 2006 91.99%
(Again, there's a lot more there, only including the top 5 because they're the closest fit.)
This blends to (SVP/GAA/VORP/nERD per gm):
2017-2018: .926/2.17/+7.13/+0.45
2018-2019: .924/2.32/+5.11/+0.33
2019-2020: .928/2.11/+8.93/+0.52
---
4. Conclusions
My first reaction to the player array was that they were a little bit bearish; I don't really consider, for example, Marty Turco to have ever been as good as Matt Murray is now. Of course, part of that is my bias as a Pens fan, another part is my likely ignorance of Turco, and so on. I suspect part of the surprising results is due to sample size; 62 NHL game is not a lot to go off of. Still, it points to the fact that while we as Pens fans are all very bullish on Matt, careers can and often do take dramatic swings (Varlmamov, Backstrom), and it would be foolish for us to assume permanency after only 62 games.
(I noted in section 2 that some of the comparables were a lot further developed in their career - such as Mike Smith 2012, who was 29 at the time. Halak 2010 was 24 and Varlamov was 25; it makes my life easier for the sake of apples-to-apples conclusions that those two are pretty close in age as well.)
Of the top comparables, Halak maintained good form but I'd say never really became elite; Varlamov fell off pretty hard. That's not to say that those results are expected for Murray, rather it's to say that they are possible, given that they're the top comparables and that as both Halak and Varlamov looked to be on the verge of being very-good-to-excellent at the time. This probably makes Halak the reasonable floor here - not counting injuries of course - as Murray's numbers likely will never look as rough as Varly's look because the 17-19 Pens will likely be much better than the 12-14 Avs and because Varly didn't have elite numbers in the AHL that Murray has.
Given that none of the stronger comparables strike me as being elite, it's hard to quantify what the ceiling is, particularly given that Murray has outpaced even the rosiest of projections every step of the way so far. I'm inclined to think that the low-end is something like a better, more consistent Luongo playing for a better team, or a Rask/Varlamov without their recent downward trends. I'm tempted to use Price as well, but given the relative weakness of the similarity (not to mention how rough Price's three years after 2011 looked), it strikes me as being less of a fit for our purposes - Murray could (and probably should) outperform Price's 2011-2013, but there's similarly weak indication that Murray can consistent deliver Price's 2013-16. For the purposes of a quick blurb, I'd call it somewhere between Luongo+ and Price as the ceiling.
(A few posters upthread mentioned Holtby as a comparable; for what it's worth, Holtby doesn't appear in the top 20 comparables for either players or projections. This is likely because Holtby's improvement from 13-14/14-15 was somewhat unique from a statistical perspective (2.85/.915 to 2.22/.923) and as such, it's generally seen as an outlier/unlikely growth curve for anyone, not just for Murray.)
One thing I meant to mention in the projections section is that in addition to having a very solid worst-case in Halak, Murray also shows a low variance/high confidence interval. This should give the Pens and you, the Pens fan, a lot of reassurance that even though you might intuitively disagree that Luongo+ is a high enough ceiling, the consistency is projected to a degree that barring injury, there are no bad outcomes; a dropoff like Varlamov or Rask is unlikely. They do happen, however, and therefore they must be accounted for even if it is difficult for us as humans and biased fans to do so.
The worst case scenario is likely that even with some regression/growing pains, Murray is a consistent top 10 goaltender in the league for the next 5-6 years - something that you'd have to stretch to say about MAF, if we're being honest - and obviously the upside is significantly above that. The likely outcome and ceiling are pretty closely clustered (again, due to low variance), profiling a yearly Vezina candidate that delivers between .925 and .930 year in and year out.
I'd be interested in revisiting this in a year, hopefully after another Cup and 50+ more starts under his belt.
Holtby is the best comparison for Murray's future success, oops I better rephrase that
Holtby is the best comparison for Murray's future statistical success with Save% and GAA. As far as clutch winning success in the playoffs you can't find anyone better then 2-0 in the finals and 6-0 in playoff series. You can't overlook the clutch part of goaltending and playing your best when all the chips are on the line.
