To quickly recap Part One of this undertaking, we expanded the scope of raw GSAx by comparing it to different contextual factors like consistency, defensive environment, etc. Now that the data points have been established, it’s time to decide how to deploy the findings.

Essentially, I wanted to put myself in the mind of an NHL general manager in need of a goalie. What type of goalie should I be looking for? To what extent should the complexion of my existing roster play a role in the type of goalie I pursue? How should the different factors we examined in Part One be weighted in the decision-making process?

Let’s try to tackle those items and come up with some best-practice theories. I want to stress the word “theory”. I didn’t unlock some secret amazing formula for goalie projection. In reality, I just looked for contextual trends to add to the analytics arsenal beyond GSAx when discussing goaltenders.

Anyway, let’s wrap up this monstrosity, shall we?

Theories on Assessment Philosophy

I’ve just spewed a lot of data in your general direction in Part One. We should probably try and piece it all together.

From what I’m seeing, analyzing and predicting netminder performance is about as difficult as I initially believed. That being said, I think that certain findings were telling, and perhaps provide a fractured blueprint of what teams should be looking for.

First and foremost, you want to see good performances over a long period. That’s obvious. The second piece should be an analysis as to whether a good, consistent goalie has done so in a myriad of different (and/or particularly difficult) defensive environments (charts above and below reposted from Part One for ease of reference)

For example, goalies who have succeeded in exclusively strong defensive environments could still be good behind a below-average xGA team, but it’s a roll of the dice as there would be no historical proof.

Lastly, you want to make sure the above results coincide with something resembling a consistent year-over-year workload (depending on the role in which you plan to utilize said goalie). That piece is critical as it helps you avoid investing in a goalie who maybe checks all the initial boxes, but has served exclusively in a backup role. Conversely, it gives general managers an idea of what size workload can be reasonably expected from a given goalie moving forward.

Jake Allen is a fantastic example of what I’m talking about. If you go by the charts I created, you could convince yourself that he’s one of the most solid, consistent goalies in the NHL. That’s true, but it’s important to note that he’s done it with a pretty small workload compared to his contemporaries in this study. Inherently, if a GM was to consider him for a starting role, he would need to consider that the change in workload could negatively impact otherwise impressive per-60 numbers.

Is it possible to find a goalie who “checks all the boxes” I’ve outlined? Sure, his name is Connor Hellebuyck. End of list. Consistently awesome and handles a heavy workload in front of volatile and bad defenses. He’s the only one from the study who is truly elite. Juuse Saros is close behind, but Hellebuyck is in a league of his own.

Apart from that, the contextual data I’ve assembled casts varying levels of doubt on pretty much every other goalie in the study. From there, you just have to decide which factors mean the most to you based on how your team is situated/constructed. Speaking of which…

Team-Building Dynamic

I spoke with our good friend Kevin (also known as ntrider825 on Twitter), about this exercise as I was going through it, and he said something that sort of inspired a theoretical debate in my brain.

We’ve got all this data, right? We know which goalies were good, which were consistent, and the types of defensive teams they played behind. How you go about ranking and targeting these goalies might not just depend on the goalie himself, but also on your status as a team.

Mock Hypothetical

Let’s do a fun hypothetical. Say you’ve got a team of skaters that you believe are capable of a deep playoff run, and your team ranks in the top half of the league in xGA rate. You’re heading into the summer with no starting goaltender on the roster. The following three goalies from the above sample are interested in joining your team (and we’ll assume you have enough cap space to sign any of them).

The options are Joonas Korpisalo, Tristan Jarry, and Jordan Binnington.

Which of them should a team like the one I’ve described pursue? Well, let’s think about them individually. Korpisalo is the only one of them to have produced a crazy GSAx bender at any point over the last four years (GSAx/60 rate of .394 just last season, which ranked 14th in the NHL). If he were to hit that mark again, it could take your team from contender to cup favorite.

Conversely, he’s also produced some abysmal results in the years prior, and if he regresses to his previous norm (average GSAx/60 of -.555 from 2019-2022), it could severely limit your team’s otherwise promising potential. It might even push you to the fringe.

On the other hand, you have Jarry and Binnington. The former has produced decent results consistently in front of strong defenses comparable to your own. Binnington, on the other hand, has been similarly effective, albeit not quite as consistent (and less so recently). That being said, he’s played in a much less favorable defensive environment throughout the entirety of the sample.

On your team, there’s a chance to unlock a new level in his play. Or maybe he’s exactly who he is regardless of how well his defense plays in front of him. Cam Talbot is an example of this type of player from the study. Lots of different defensive environments and, same lackluster results. Binnington is, of course, much younger and full of potential. He’s also won a Stanley Cup, so he’s seen what that requires.

Decision Time

Who do you choose? Well, right away I’d eliminate Korpisalo from the list. While there’s a chance he could replicate his 2022-23 performance, there’s a whole lot of bad that preceded it. It’s too great a risk for a team otherwise primed to make a run. We’ve seen this movie before. The voodoo monster almost always eats the unsuspecting GM.

Now it’s down to Jarry and Binnington. If this mythical GM believes that his skater depth (particularly on defense) is among the best in the NHL, Jarry is perhaps the safer option. You know what you’re getting, and the results are almost always a tick or two above average. A goalie who isn’t going to elevate your group all that much, but he probably won’t hurt you either.

Now, if you feel that your skater group is close to the top, but perhaps not quite there, the risk associated with Binnington might be a worthwhile pursuit. If he shows improvement on a better team, you look like a genius. Based on what we know from the study, that’s a big “if” (i.e. team xGA and GSAx don’t always correlate).

It’s interesting. If it were me, I’d probably choose Binnington in this scenario, given all of the information I’ve collected (but it’s very close). It does bring up the question of risk mitigation and how willing a GM might be to take a chance on a less established netminder under the right circumstances. That being said, give me the guy who has fared similarly (albeit a bit worse) in a significantly more difficult defensive environment.

Luukkonen Comparable?

I’d be remiss if I went through all of this and didn’t tie it back to the Sabres in some way.

As I’ve said countless times, I’m wrong about goalies far more than any other position. It’s not even close. I’m not saying that anything I’ve outlined here will change that, but perhaps it will help me establish and justify the extent to which I support a prospective goaltending option (or vice-versa) for the Sabres moving forward.

As part of this entire exercise, I was hoping to find a goalie or two who I felt projected closely to Ukko-Pekka Luukkonen. Since he doesn’t have nearly enough of a year-over-year sample to pool from, this next bit is part data, part gut feeling.

It seems to me that Luukkonen is the type of goalie who tends to jive with his defense. On nights when the Sabres’ defense struggles, so does he. On nights when they’re average or better, however, he’s starting to shine.

Worst case, he’s similar to Elvis Merzlikins. A volatile goalie who is more directly influenced by the defense in front of him than average, but is still largely bad. Best case, he’s a Darcy Kuemper type who is influenced by his defense to a degree but can start to give you great results if said defense is average or better.

So far this season, he’s been closer to a Kuemper than a Merzlikins. That’s good news. It’s a wide spectrum of possibilities, but Luukkonen is a young goalie with a small and fractured sample. Tough to directly apply data from this study, hence the gut-feeling factor. Stay tuned.

Closing Thoughts

While this exercise certainly didn’t unlock the mystery that is goaltender randomness, I believe it helped add valuable contextual data points to consider in addition to raw GSAx. The process will require additional refining as we move forward, but I think the point of volatility and environment are factors that probably don’t get enough consideration (or perhaps their values are misattributed).

While I don’t know if I’ll have the ambition or energy to refine the results once the 2023-24 season comes to a close, it would be interesting to revisit. This season is shaping up to be a doozie in terms of GSAx volatility, and I’ll be curious to see how it impacts some of the ideas/practices I’ve theorized above.

At the very least, I hope this allows folks an avenue or framework for things to look at beyond raw GSAx when analyzing goaltenders moving forward. Maybe it’s all useless. A handful of players from this very study are already pacing out to throw off their previous trends in 2023-24. Maybe they regress toward the mean, or maybe these monsters are just completely volatile and inherently risky, save for an elite few.

Eat Arby’s.

Link to Part One Here

Advanced Metrics courtesy of Evolving Hockey and Natural Stat Trick

Charts courtesy of Expected Buffalo and Evolving Hockey

Photo Credit: Ethan Miller/Getty Images