Goalies are not Voodoo

The tools used to measure goalie performance are simply not as good as those used on skaters. While we now acknowledge that Goals Against Average tells very little about the goalie and more about the team, we still cling onto Save Percentage (SV%) as the only port in the storm. Even in its refined forms like the War on Ice Adjusted SV%, it has always suffered from two problems: the measure does not separate goalie skill from team skill well enough, and by the time you have enough data to make a reliable judgement on a goalie, you have just witnessed his peak years go by.

Scouting and other statistical measures based on non-public data are the tools teams have at their disposal to add on to the imperfect measures we can all use. And it seems from an outsider point of view, that scouting is what takes precedence when teams value one goalie against another.

Goalies are Gambles

All hockey is about gambling and risking for reward. Smart teams amass the most knowledge to mitigate the risk and to make better choices, but they are still gambling. A wrong guess on a forward can often be saved by playing him down the lineup or in the AHL. The most you lose is money or cap space. A wrong guess on a goalie might find you suddenly getting a much higher draft pick than you expected at the end of the year. They are hard to hide when your bet goes sour, and even harder to offload if they really blow up in your face.

I see the goalie trades the Leafs have made as a bet on Frederik Andersen over James Reimer and Jonathan Bernier both. What tools they used to make that bet and what attitudes or beliefs influenced them is fodder for fiction. And only time can truly tell us if it was a good bet or not. Although arguing over it now is a tried and true pastime, and I encourage you to do that all you like. Just count me out.

Andersen is here. The bet is down. Now the issue is: how to hedge it.

The Leafs spent $2.25 million in cash and cap hit to hedge the Zaitsev bet. Roman Polak has a few other uses to the team per their recent comments—Mike Babcock likes him for his commitment to hard work in the weight room and for his general approach to the game, for example. But he is fundamentally an insurance policy.

Lou Lamoriello stated that he would find a backup for Andersen by training camp, and Babcock reiterated that, saying they would get a real pro to work behind their new starter. That sounds like serious hedging to me. And I'm relieved.

Before the second half of the Andersen deal, when it was just Anaheim selling off their starter and putting all their money down on John Gibson, I thought the Ducks were the ones taking a big and foolish gamble. Now that the other shoe has dropped, the Leafs are sitting on the high risk play. And I hope and expect them to do a lot more than throw a couple of dollars at the cheapest backup. Roman Polak speaks to their willingness to have a meaningful plan B on a risky play, and to pay for it.

Who though? Who can they get who is a pro who can step in if the Andersen bet fizzles?

The Options

On the team, the only NHL experienced goalie behind Andersen is Garret Sparks and his short tenure was a disaster.

In free agency, there is not much to pick from. There are a handful of NHL experienced "name" goalies: Ben Scrivens, Karri Ramo, Jhonas Enroth and Anders Lindback are the only ones not well over 30 and at the end of their careers. Scrivens is rumoured to be ready to sign in the KHL while Ramo and Lindback are on the low side of league average, so that leaves Enroth.

On other teams, in positions of second or third goalie where there is a potential trade to be made, I found three possibilities.

Thomas Greiss is the second of four goalies on the Islanders roster. They likely have him and Jean-Francois Berube signed as insurance against Jaroslav Halak missing games again, while Christopher Gibson will play in the AHL. The play may be to move Greiss, a pending UFA, at the deadline, but they might be persuaded to do that now.

Antti Niemi is the overpaid second half of the ineffective Dallas Stars platoon. Dallas would have to make it worth someone's while to take on his salary which runs for two years. The sweetener might be worth it if he is a good enough to meet the needs of the Leafs.

Reto Berra is the third of three goalies on the Panthers roster, but with the expected delay in Roberto Luongo starting the season, acquiring him would require waiting and starting the year with Sparks, so he would need to be very much worth the wait.

Let's have a look at these guys, as well as Andersen, and see what we think:

Player Season Team GP TOI SA Sv% xGA GA xGA-GA xGA-GA/60
FREDERIK ANDERSEN 2013-2016 ANA 125 5382.44 2518.36 92.54 212.81 190.88 21.93 0.244
GARRET SPARKS 2015-2016 TOR 17 753.83 378.98 89.86 29.28 37.26 -7.98 -0.635
JHONAS ENROTH 2009-2016 BUF/DAL/L.A 147 6343.67 3383.03 92.43 280.46 250.69 29.77 0.282
THOMAS GREISS 2007-2016 PIT/ARI/S.J/NYI 129 5399.92 2592.19 92.94 207.44 183.05 24.39 0.271
ANTTI NIEMI 2008-2016 S.J/CHI/DAL 386 17480.59 8176.67 92.43 657.44 624.70 32.74 0.112
RETO BERRA 2013-2016 COL/CGY 64 2564.86 1308.43 92.06 104.45 102.55 1.90 0.044

This is Corsica Hockey's goalie data, with a few things winnowed out to simplify the table. It is score and venue adjusted and includes everyone's full career.

I added two calculations: Expected Goals Against - Goals Against as well as that number rated out over 60 minutes.

Corsica creates their expected goals numbers from shot type and location data that is available from the NHL game sheets. They weight the factors, calculate the likelihood of an average goalie saving that shot, and determine if it should be expected to be a goal or not. Shooter skill is not factored in. The in depth explanation begins here.

Using this for goalies is, in a sense, sneaking up sideways on Adjusted SV%, which uses similar weightings to give higher or lower value to saves of greater or lesser difficulty. But it also erases the issue of high volume goalies having a higher SV% just because that's how math works.

I took the difference between expected and actual goals which could be called Goals Saved Above Expectation. If you are familiar with Goals Saved Above Average, this is merely a weighted for shot quality version of that. A negative number is bad. The bigger the positive the better, but of course all of that is affected by how much each goalie played. Hence, the rate stat.

N.B. there is ongoing research about this type of stat to determine its predictive value. So far the results are mixed. It works in some forms for skaters, it does not so far have a strong repeatability in goalies. One likely problem is that the definition of expectation or average save is not good enough. Another is the perennial low amount of data issue. This is presented here as another point of view on historical performance along with SV%. Neither has been proven to predict future performance very well.

So what do we see here? Obviously Sparks did not perform up to average goalie standards. His SV% shows that, so does his negative number in Saves Above Expectation. I'm calling this SAE from now on.

Andersen, Enroth and Greiss all have similar SAE. They have similar SV% too. They all seem like they have performed above average.

Niemi, who has a high SV% too, has a very high gross number of SAE, but when rated out for his large number of minutes, it dwindles to very near average.

Berra looks like a league average guy from all measures.

The choice, assuming there isn't some amazing, professional, experienced backup goalie we haven't thought of out there, seems to be between prying Greiss away from the Islanders or signing Enroth.

Greiss became the Islanders starter this season by accident, and he performed very well. He would cost in assets to acquire more than money as his contract is $1.5 million, and he is UFA next year.

Enroth is an unsigned UFA, who seems to have never caught on in the minds of the hockey world as a good goalie. He picked up a bad smell after years of backstopping terrible Buffalo Sabres teams and was ultimately traded by them for being too good for a tanking team.

When he and Lindback both split the 2014/15 year between Buffalo and Dallas, they ended with similar SV%, but again Enroth's expected goals against vs. actual was much more positive. Literally positive, to Lindback's negative. If ever there was a pair of players to make a misleading SV% graph with, surely those two would do, although I can think of others.

In a very small number of games in LA, Enroth looked very good.

The unknown factor is: how safe do the Leafs think their Andersen bet is, and how much do they want to spend to hedge it?

They could conceivably get something very nice by taking Niemi and living with his Bernier-like performance and salary. They would have to give up something for Greiss while all Enroth costs is the money and cap space needed to sign him.

Or maybe they are much more clever than I am, and they have their eye on someone else, someone they have scouted for as long as they did Andersen. It's not like this is a new problem after all; they've had months to work on it. And they scouted Andersen for most of last season, so the issue of a backup has to have been in their minds for all of that time too.

Who do you want as the backup?

Pick your plan B

Garret Sparks173
Jhonas Enroth1260
Thomas Greiss690
Antti Niemi193
Reto Berra27
What was wrong with Bernier?206
None of those280