Arbitrary Endpoints, Career Numbers, and the Future of Jacob Markstrom

(As an introductory note, it's been a while since I posted here, but a lot has happened in the interim, not just on the ice during the 2014-15 season but especially off of it, with the Summer of Analytics, the emergence of new statistical resources, and some interesting perspectives on goaltending from Steve Valiquette, Chris Boyle, and others. On top of all that, I was also inspired by attending the Ottawa Analytics Conference and meeting many of the people I have worked with and interacted with online.  All that means I'm planning to get back into hockey analytics, which for the time being will include contributing to the Hockey Prospectus Annual and blogging in this space.  Although I was pretty fortunate in having my last post here be on a topic that makes me look pretty good in hindsight (Dubnyk vs. Scrivens), it's long past time to starting pushing that one down the page.)

One of the things I have been thinking about lately, and something I wanted to bring up in my first post back, is the issue of career stats and arbitrary endpoints, to borrow a term used by ESPN sabermetrician Keith Law.  Here's the description from Law's glossary:

#arbitraryendpoints: Also known as cherry-picking, this means choosing one or both endpoints on a series of games to try to analyze a player. I’ve argued that it’s not arbitrary if the endpoint is tied to something specific, like a change in mechanics, an injury, or a recall from the minors, but even so, it’s always dangerous to throw out any data when you want to draw a conclusion.

I definitely agree that it is dangerous to throw out data and that it should never be done for the purposes of supporting an already-reached conclusion.  However, I also agree that it is not always arbitrary to look at splits and segments of data rather than relying on the complete sample if there is good reason to expect that some of the data is not representative of an athlete's true talent level.

There seems to be an increasing trend on Hockey Twitter for people to simply pull up a goalie's career save percentage or career EV SV% and use that as the final verdict on their talent level.  This can certainly be appropriate some of the time, perhaps even most of the time.  I still think that goaltending is above all else a results business, and that statistical measurements remain very powerful methods of evaluating performance.  That's why save percentage over a large sample size is usually a good proxy for a goalie's talent level.  However it is still only a proxy, and the three points mentioned by Law do also apply to goaltenders.  In the same way that Law might discount a pitcher's performance in his first year back from an arm injury, we might have reason to believe that a goalie could be deviating from his historical average because he is not yet at 100% health after coming back from some time off or has made some changes in his game after working with a new goalie coach.

On top of those factors there are others more specific to hockey goalies, such as team effects/shot quality, situational performance (e.g. EV vs. PK), home scorer bias/road performance, usage, etc.  These factors generally do not have a major impact on stats, but margins are so slim in goaltending that even a slight advantage or disadvantage can have an effect on the rankings.  Overall, I think sometimes insight can be missed by looking only at the big picture, and in those cases it is appropriate to take a deeper look.  And I'll start doing that by focusing a magnifying glass on the short but interesting pro career of Jacob Markstrom.

Scrivens vs. Dubnyk

Just 1309 shots ago, Devan Dubnyk was a three-time Team Canada representative at the world championships making $3.75 million a year and seen as the future of goaltending in Edmonton, while Ben Scrivens was a backup goalie making $612K that started just 2 of his team's last 19 games before being included as the goalie going the other way in the Bernier deal.

How things change.  Dubnyk has struggled in 2013-14 (.894) while Scrivens has thrived on arrival in L.A. (.931), and as a result Edmonton's management figured it was worth pulling the trigger on two separate trades to effectively make that goalie swap.

I think there are might be a couple of legitimate reasons to prefer Ben Scrivens over Devan Dubnyk if one was forced to choose between the two of them (a cheaper cap hit being the main one), but I don't know why anybody would care that much.  It seems to be an obvious overreaction by the Oilers to their goaltending issues.

Both Dubnyk and Scrivens are 27 years old and pending UFAs.  Scrivens has a .917 career save percentage to Dubnyk's .910, but Dubnyk has almost 4 times the sample size at 5079 SA compared to 1365.  Over the past 4 seasons, Dubnyk's situational numbers aren't too far off Scrivens' (.920 EV, .876 PK vs. .923 EV, .891 PK), and that's despite playing on an Oilers team that averaged a 68 point pace over that span (including 34-70-15, .902 without Dubnyk in net).  Scrivens had better minor league numbers, but as a collegian he was more developed when he hit the pros whereas Dubnyk came from the Dub at the age of 20.  Dubnyk was a first round pick, while Scrivens was undrafted, and Dubnyk would likely have been considered ahead of Scrivens in development throughout their entire careers until a couple months ago.  I'm not sure half a season is enough to make a major correction in that relative ranking.

I find it very interesting that Nashville was the team acquiring Dubnyk, since the team's continued employment of Mitch Korn and track record of goalie development makes me generally trust their instincts on goaltending.  Dubnyk is going to have a good opportunity to turn things around with the Predators.

Even if the Oilers are right, though, what do they really gain by this move?  There are just 33 games left in another lost season, where the only real goal left to accomplish is to continue sinking faster than the Sabres and Flames and thereby gain the inside track on yet another first overall pick.  Ilya Bryzgalov is also still in the mix, and it seems likely he will get his share of starts, making acquiring a goaltending upgrade even less valuable to the team.

Assume that Oiler goalies other than Bryzgalov get 500 shots the rest of the way.  If Scrivens keeps up his .931, he'll allow about 35 goals, or 18 fewer than Dubnyk would have if he held steady on his seasonal pace over the same workload.  That's a sizable difference, equivalent to three wins in the standings, yet how likely is it that both goalies don't regress heavily towards their career averages, and what's the advantage in finishing 28th instead of 29th or 30th?  Even if they are philosophically opposed to tanking, if Edmonton is serious about handing the reigns over to Scrivens it would seem that they would have an awfully good chance of signing him this summer.  The Kings certainly aren't going to be much competition for any Edmonton offer given that they have their starting goalie locked up through 2023, and with Dubnyk's $3.5 million coming off the books Edmonton would likely have more than enough cash to outbid any other teams in Scrivens' pay range.  It appears they must be intending to give Scrivens a tryout to see if he merits a contract in July, but even if they do that they're making a decision from a fairly small sample size.

Then again, panicking over small samples seems to be something of an Oiler trademark.  If any team should be patient, you'd think it would be a team in development mode, yet the Oilers have been the king of goalie overreactions this season:

1. Signing Jason LaBarbera as the backup goalie, then dumping him to Chicago after only 146 shots against.

2. Bringing up minor-leaguer Richard Bachman and giving him three straight starts because he played pretty well in the first one.

3. Signing Ilya Bryzgalov, a goalie who nobody else wanted, to a one-year deal after Bachman got injured and Dubnyk and LaBarbera continued to struggle.

4. Trading Dubnyk and acquiring Scrivens.

The other big issue is the cost Edmonton paid to make the move.  They gave up a 3rd round draft pick (which given the Oilers' poor play will be a very high 3rd), plus they retained $1.75 million in salary on Dubnyk, plus they took on the contract of Matt Hendricks, who is signed for $1.85 million for the next three seasons.  Hendricks is 32 years old and has a -0.3 GVT this season and just 21 points in his last 170 games. Given his age and recent performance, he's very likely to be a replacement level player at best over the next three seasons, which means the Oilers are probably wasting about $4 million compared to signing an equivalent guy to the league minimum to plug in on their fourth line.  The Oilers did save the difference in salary between Dubnyk and Scrivens over the rest of the season (~$1.2 million), but with the retained salary and likely overpay on Hendricks it looks to me like they may have paid in the range of $4 million plus a 3rd rounder for 33 games of a goalie that they could have had for free this summer.  All in all, a perplexing move for Edmonton.

Goalie Utilization at the Olympics

From the reports I heard, Mike Babcock is looking to pick one goalie as his starter and run with him, rather than auditioning several goalies out during the preliminary rounds.  I think that makes sense, because trying to find the hot goalie is probably little more than shooting in the dark to begin with, and it's better to just throw your chips in on the best guy.

EV SV% since 2010-11:
Luongo .928
Price .926
Smith .926

EV SV% since 2012-13
Price .927
Smith .923
Luongo .922

There's probably not that much between Canada's goalies.  Chris Boyle has made the data-supported argument that Price has faced more difficult shots, which I certainly find interesting, and with his slight advantage in recent play and his strong form this year I do think Price should have the inside track on being the guy in Sochi.

That said, the compressed Olympic schedule means that it likely would not be sensible to start one goalie all the way through.  Canada plays 3 games in 4 days in the preliminary round, and if they don't earn a first round bye and have to go through the qualification game, then they would be looking at another 3 games in 4 days just to get through to the semifinals.  A semifinal loss means a bronze medal game the day after, while at least the gold medal game participants get a much-needed day of rest prior to the final showdown on Feb. 23.

Given the data on goalies making back-to-back starts, and considering that most countries have at least two quality netminders on their rosters, the argument could be made that few if any countries should be starting the same goalie in back-to-back games.  Canada should be in good shape no matter what they do against Norway and Austria, since you'd expect those to be easy wins and likely a light workload for their goaltender regardless, but it would probably be a good plan to split those starts.  And while it would be an extremely gutsy decision by any coach who elects to do so, if Canada ever got to a qualification game against one of those teams or another weak sister nation like Latvia or Slovenia, there would at least be an argument for the team to start their backup goalie and save the real starter for the quarterfinal the next day.

I really doubt any team would actually do this because if it ever backfired the coach would be raked over the coals, and you'd probably create all kinds of controversy by switching back and forth between goalies because everyone wearing a media badge loves to pounce all over that kind of manufactured drama.  No teams tried this at the last Olympics, as all 4 qualification game winners put the same netminder right back between the pipes for the quarterfinals.  That said, this strategy would make the most sense for a powerhouse team like Canada, because it would seem to have the least risk.  Goaltending didn't matter in the 8-2 playoff win over Germany in Vancouver, it seems almost certain that Canada would have ran away with  the game if either Brodeur or Fleury were tapped ahead of Luongo.  As it happens goaltending didn't really matter against Russia either, because of Canada's outstanding offensive display in scoring 7 goals, but it's not hard to envision an alternate scenario where Luongo's 25/28 would be not good enough to advance.  Don't even bother try to imagine one, just go back four years prior to when Brodeur's 31/33 QF against the same opponent resulted in a disappointing early exit for Canada.

I'll be watching Olympic goaltending decisions with interest, both in terms of why plays and when.  I'm expecting pretty conventional decision-making all around, but if anybody decides to think outside of the box I'll be impressed, whether it works or not.

One Shining Moment

I used to be much more of a baseball fan than I am right now, but on occasion I can get suckered in binge reading Joe Posnanski or someone like that, particularly in the middle of one of the sport's classic debates between sabermetricians vs. traditional media members.

Jack Morris' Hall of Fame candidacy has been one of those types of affairs, a 15 year debate about the value of a durable pitcher with a few shining postseason heroics but fairly underwhelming overall numbers. Plenty of different rationales have been offered by the pro-Morris camp, but I think that as much as any recent athlete I can think of, Morris is simply defined by a few big moments (most notably game 7 of the 1991 World Series), and the residual impact of those highlights trump all other factors in the minds of those who tend to rank players based on their gut feel as opposed to a detailed comparative analysis of their performance record.  Many of the arguments in favour of enshrining Morris seem to be people reaching for justification to support their already decided position.  That can of course be fairly criticized for being poor analysis, but I don't think you are going to convince many people purely by calling out their analysis, especially when they are being largely driven by subjective perceptions.

I think the power of the shining moment can be just as significant in hockey, perhaps even more so given the overwhelming importance of the Stanley Cup playoffs.  Just as with pitchers in baseball, winning or losing tends to stick with goalies in that it is easier to recall the successful plays of the winner and the negative ones of the loser.

One of my personal favourite examples of this is Marc-Andre Fleury in the 2009 playoffs.  Many hockey fans called him clutch and said he made the big save at the right time, because they could easily recall his stop on Lidstrom with time running out in the Finals in game 7, along with his breakaway save on Ovechkin and a few stops against the Flyers in round one.  But the rest of Fleury's playoffs were pretty ordinary (.908).  Nobody is able to fully recall even a small sample of shots like the 686 Fleury faced that postseason, it's going to inevitably filtered down to a few memorable ones and if the ones that stick are saves a goaltender will come off looking good.

I also think it people took longer to figure out Ondrej Pavelec was actually a pretty mediocre goalie because of his propensity to make the big highlight reel saves.  Fans would be travelling home from a game thinking of that one he snatched out of mid-air and not about the fact that he let in 4 out of the other 29 shots against, but it is how many you let in out of how many chances that really matters.

On the other hand, negative perceptions can be similar.  Tony Esposito seems to be remembered much less fondly by many older hockey fans than his overall record would suggest, and it doesn't take long at all in any discussion of his merits before somebody brings up Jacques Lemaire scoring on him from center ice in game 7 of the Finals.  Tommy Salo wasn't an all-time great, but he was a solid NHLer with an impressive international carer even though it remains awfully hard to think of him as even a competent netminder at this point because of the taint of that one Belarus goal influencing our collective memories.  I think that is how bad-team goalies tend to be forgotten and dismissed, it was always curious to me when someone like Allan Bester was treated as a punchline despite fairly impressive stats, but if you can't think of a goalie making a big save or winning something then you're probably going to think of some bad goal or another.  No matter how good they are, no goalie ever comes off looking good in a highlight reel of their goals against.

I believe countering this effect is one of the most important uses of the statistical record in evaluating athletes.  There are some legitimate quibbles about statistical recordkeeping and the reliability of various metrics, but at the same time all observers also have biases and faulty memory, and it's well worth having an extra check to ensure that the events our brains remember did actually happen in the way that we claim they did.  Maybe a player was clutch and "knew how to win", or maybe certain events stick in the memory more strongly than others.  The burden of proof should always be on the former position, in my opinion.

New Jersey Home/Road Split: Undercounting or Style of Play?

Home/road goalie splits for the New Jersey Devils this season:

Home:  .942 Save %, 20.98 SA/60
Road:  .910 Save %, 27.12 SA/60

A difference of 6 shots per 60 is a massive gap, and likely is at least partly due to some small sample size randomness.  However, New Jersey has long been suspected of undercounting shots at their home rink.  What is curious, however, is that the team save percentage is much higher at home than on the road, which is not what would be expected if it was purely a scorer counting issue (see, for example, Martin Brodeur's career splits, which work out to .911 at home vs. .914 on the road, numbers more consistent with a theory of systemic undercounting).

Extra Skater gives possession data at the team level, and those stats indicate that New Jersey is the lowest-event team in the league no matter where their games take place:

Most Low Event Home Teams (Combined Fenwick events per 60 minutes):
1. New Jersey, 58.2
2. Minnesota, 70.9
3. Buffalo, 73.6
4. Columbus, 74.0
5. Calgary, 75.4

Most Low Event Road Teams (Combined Fenwick events per 60):
1. New Jersey, 71.3
2. Edmonton, 74.7
3. Pittsburgh, 74.7
4. Minnesota, 74.7
5. Columbus, 76.9

It certainly seems possible there is some undercounting still going on in New Jersey since they come in so far below all other teams in home Fenwick events, although it should be noted that there were two other teams that had a similar split in terms of a much lower game pace at home compared to on the road.  The Devils' differential of 13.1 only narrowly edged out Ottawa's 12.3 and Buffalo's 11.8.

It is also interesting that the Sabres and Senators have very different goaltending splits during 5 on 5 play (although these are very tiny sample sizes):

Buffalo:  .906 at home, .934 on the road
Ottawa:  .911 at home, .940 on the road
New Jersey:  .931 at home, .911 on the road

The two most common explanations for a team giving up far more shots on the road would seem to be either a difference in the way shots are counted or a team trying to lower average shot quality by giving up more ice from the less dangerous areas of the defensive zone.  Either way, what you expect to see is that the team's goalies do a little bit better on the road than expected, which is again why the Devils' splits seem abnormal.

Looking at the overall population, there is some correlation between pace and save percentage in 2013-14, although it should be noted that the data shows little relationship for 2012-13.  The teams with notably different home/road splits are also quite different from 2012-13 (New Jersey had a difference of just 2.3, albeit in the same direction), with only Carolina and Dallas coming in as substantial outliers over both seasons (both of them playing with a higher pace at home than on the road).

More evidence is needed to show whether a team's pace has any kind of sustained impact on its percentages (in general it probably doesn't), but while they are no longer using the infamous neutral zone trap, it is clear that the New Jersey Devils are focusing more than any team at suppressing shots and chances at both ends of the rink.

Why Canada Doesn't Have a Goaltending Problem at Sochi (If You Believe in Large Sample Sizes)

With the Sochi Olympics lurking on the horizon, the early season fortunes of goaltenders have caused a lot of glee or consternation among hockey fans around the world, depending on how their nation's puckstoppers have come out of the gate.  With respect to goaltending, however, it is important to remember that talent is revealed in the large sample size (see here for one of the definitive takes on that theme).  There has been a lot written lately about how Canada no longer has top-level elite goaltending, but in my opinion that was an overreaction to a shortened lockout season where it was purely coincidence that the country's best seemingly all had off-years.

It's easy to get distracted by the flashy new models, especially when they are off to scorching starts, but looking at only the established goalies with a track record of success Canada simply does not come up short at all relative to other nations.

Goalies with at least 4000 SA since 2010-11 (ranked by save percentage):

1. Henrik Lundqvist: .925 on 5313 SA
2. Pekka Rinne: .922 on 5388 SA
3. Roberto Luongo: .921 on 4359 SA
4. Mike Smith: .920 on 4217 SA
5. Carey Price: .919 on 5624 SA
6. Kari Lehtonen: .919 on 5286 SA
7. Antti Niemi: .918 on 5290 SA
8. Jonathan Quick: .918 on 4771 SA
9. Cam Ward: .918 on 5211 SA
10. Craig Anderson: .917 on 4587 SA
11. Marc-Andre Fleury: .916 on 4820 SA
12. Ryan Miller: .916 on 5590 SA
13. Jimmy Howard: .915 on 4923 SA
14. Jonas Hiller: .914 on 4493 SA
15. Niklas Backstrom: .914 on 4064 SA

The three goalies that seem likely to be named to Canada's roster are all in the top 5.  Luongo and Smith are both off to solid starts this season, while Price has been outstanding.  The three have them have combined for a .922 save percentage this season, which is very much in line with their historical numbers.  In fact, the trio have been very consistent, with the exception of one lockout-shortened 48 game season:

Luongo/Price/Smith Combined:

2013-14:  .922
2012-13:  .907
2011-12:  .922
2010-11:  .922

When framed like that, it looks pretty blatant that everyone was panicking over an obvious outlier.  The problem remains of picking one guy to be the starter, since when it comes to crunch time you can't combine stats and play with an average of three guys.  Even elite goalies have off games, and in short tournaments it's always better to have the hot hand rather than the superior talent (although the best goalies are also the ones with the best chance of running hot or, perhaps equally important, avoiding a very costly slump).  Overall, though, there is no question that Canada has proven goaltending talent that is at least competitive with the other top nations.

Obviously the parameters of the above list rule out a number of goalies who are likely in the running for Olympic spots (netminders such as Rask, Varlamov, Bobrovsky, Crawford, Schneider, Lehner, Holtby, Bernier, Reimer, etc.), and I'm certainly not saying that none of those youngsters deserve spots, but it is much more difficult to identify their true talent level because of the relatively small sample size they have behind them in their shorter careers.  Russia, in particular, seems to be facing a tough choice between Semyon Varlamov and Sergei Bobrovsky, both of whom narrowly missed the cutoff for this list but would have come in on the lower end anyway at .914 and .916 respectively, and the U.S. and Finland both have a number of quality contenders and limited spots, which will force some interesting roster decisions.  At the end of the day, though, I expect that this Olympics will be similar to the last one in that the top nations will all have capable goaltending and it will come down to which teams play the best overall that determines the final result.

Why the Oilers Don't Need Bryzgalov

The Edmonton Oilers didn't need to sign Ilya Bryzgalov, because goaltending isn't going to be make the slightest bit of difference for them this season.  And that's even assuming he is an upgrade on what the team currently has on the roster.

It is true that the Oilers have an abysmal team rate of .881 so far this season, with both incumbent starter Devan Dubnyk (.881) and backup Jason LaBarbera (.858) off to horrific starts.  As a result, goaltending has been the scapegoat in some corners for the team's early struggles.  While it is obvious that getting more saves would have led to more points in the standings, the reality is that the Oilers are a terrible team in a stacked division that probably wouldn't make the playoffs if they had prime Dominik Hasek tending the twine.

The bad news is that Edmonton currently has a a 44.9% score-tied Fenwick, indicating that they are being routinely dominated by their opposition.  The even worse news is that so far the team has played the majority of its schedule against the inferior Eastern Conference, going just 0-5-1 so far in games against Western Conference opponents.  The top five teams in the Pacific Division currently all have records of .600 or better, and are a combined 40-14-7 against the rest of the league.  If Metropolitan Division teams can run the Oilers off the ice, then it's going to be no contest against the heavyweights of the West.

Even though it is early, and some teams records are skewed based on their number of inter vs. intra-conference games, the wheat seems to already be separating out from the chaff in the Western Conference.  Unless you're a big believer in Winnipeg or Dallas, it looks like nine teams will probably be challenging for the eight spots, and the current 8th place team in points percentage is Los Angeles at .639 (and while the Kings just lost Jon Quick for a while, they have played a tough schedule made up of primarily Western Conference opponents).

For the Oilers to be in contention for a spot, they'd probably have to have a goal differential of something approaching the Kings' +8, a difficult proposition considering Edmonton ranks second-last in the Western Conference at 2.35 goals per game.  That implies a team save percentage of .930+, and that's not even taking into account the fact that the Oilers' shot ratio is skewed because they have spent so much time trailing (if their goaltending held them in games longer, the other team would be pouring even more rubber on net and pushing the required break-even save percentage even higher).

Look at Dallas and Montreal, two teams that have very similar offensive results to the Oilers so far, while also possessing above 50% score-tied Fenwick scores and elite goalies off to outstanding starts (Lehtonen .932, Price .933).  Despite all that, both teams currently sit 5th in their divisions, with Montreal slightly better off because they are currently in position to grab the wild card crossover spot.

Ilya Bryzgalov's career high seasonal save percentage is .921.  That came three years ago, when he was 30.  Bryzgalov does not seem like the type of the goalie who would age well, given that he relies more on a blocking style and does not have top-drawer athleticism and lateral movement.  He was shredded in the KHL during the lockout, put up a .905 over the past two seasons under the microscope in Philly, and will likely take a bit to play himself into game shape after spending most of this season waiting for the phone to ring.  In short, the best case scenario for what Bryzgalov can offer the Oilers isn't even remotely close to the elite range that they're going to need to turn things around, and even if Bryzgalov could hypothetically put up .930 goaltending the team is still sunk because of their slow start and tough division and questionable backup goaltending (although I do expect Dubnyk to re-discover his game as the season goes on).

Edmonton's season is already lost, and it's because their team is awful in all areas, not just goaltending.  Giving a declining goalie a look-see won't save anything, and if the goal is to look to the future I don't think a bad stretch of 14 games changes the fact that the 27-year old Dubnyk would seem to be a much better bet than the 33-year old Bryzgalov.  I don't see the urgency that prompted the move.  If the Oilers really wanted to solve their long-term goalie problems they should have either taken a run at one of the elite goalies who might be available through trade (someone like Ryan Miller) or waited until July to try to sign somebody.  Signing Bryzgalov is too little too late, and it would be better for Edmonton to get more information on Devan Dubnyk and give him a chance to re-prove himself than to hopefully lose slightly less often with another goalie who likely won't figure into the future of the franchise anyway.