Goalies NHLe and League Equivalents

Deciphering Voodoo stuff

In the hockey analytics world, there is a whole bunch of uncertainty, exploratory questions and assumptions that are yet to be fully scientifically proven… before the game changes again.

And then there are goalies.

“Goalies are voodoo” is now a common saying in hockey. Said with a nervous smile and a shoulders shrug to admit we don’t understand a lot about goalies.

And to be fair, in the NHL, goalies have huge up and down seasons. A rookie can lead his team to the Stanley Cup more often than not when the best netminders out there can retire without ever lifting the trophy. Basically, their performances can vary so much that most teams these days rather have two above average goalies instead of one clear starter and a back-up.

So is there still a way to predict goalies? Yes. We can still try.

Can we compare leagues like we do for players? Yes, we can still try.

There are only a few issues to deal with in order to create NHLe numbers for goalies.

NHLe? You surely have read our article on rethinking NHLe for players, or other similar works.

But, to cut it short, NHLe coefficients are drawn by using players travelling between two leagues as a proxy for assessing each league respective strength. If players scored on average 1 point per game in league A and then 0.5 point per game in league B, the coefficient between league A and B is of 0.5 as their performance is expected to diminish by half.

The first issue with goalies is that not many of them change leagues, which gives us a lot of small sample sizes even going back 15 years in our database.

The second issue is, which stat do we want to use? Saving% is based on somewhat reliable data depending on each league. And does not use a lot of context to explain performances.

Once again, we’ll go for Win Shares instead. Goalies Win Shares use the number of goals allowed, shots faced, minutes played, relative to the league average of shots faced and goals allowed per 60 minutes. That way, playing in a very good or very bad team influences your Win Shares, as it should be. Or playing in a high or low scoring league as well.

We can also calculate it for most leagues, which allowed us to cover 50 leagues like we do for players. In the end, we established the average Win Shares per game for goalies in league A and look at what their Win Shares per game were in league B, ending with a coefficient (Win Shares per game in new league A / Win Shares per game in previous league B) superior to 1 if they performed better than in league B, and below 1 if they did worse. 

Decisions and Assumptions

1/ We are using the same system than for players. So please refer to that previous article linked above, so I’ll keep it short here.

2/ Like for our other works, we are using goalies moving between two leagues in adjacent seasons for reasons explained in the player NHLe’s article.

3/ We will be using a method of Selective Network.

• We chose to only be moving forward, so when the coefficient between two leagues is inferior to 1, indicating that we are progressing from an inferior league to a better league.

• We set up a cut at a minimum of 250 minutes played in a league per season to make a goalie eligible.

• We tried to only use links between leagues that have at least a sample size of 10 goalies having travelled that way. But it was not always possible (more on that later).

Selective Network NHLe

Selective means establishing a pyramid system building up the path to the NHL, which implies making subjective choices at some point so the path makes the most sense.

• Unlike for players, the second level of the pyramid was clear, the AHL was the independent final steps before the NHL. Then the KHL was the third level.

• Then we were able to plug all of SHL, Liiga, Czech Extraliga, VHL and ECHL, all of them going through some or all of KHL, AHL and NHL.

So, for example, the SHL have two paths in their networks, going through the AHL or KHL, the link going directly to the NHL having only 9 occurrences it was not included. We then weighted each path according to how many goalies went through it and calculated the final NHLe.

For the SHL, 61% of goalies progressing to a better league went to the AHL with a very tight 0.99 coefficient, and 39% went to the KHL. The calculation for each path is simply to multiply its components with each others:

                   SHL > KHL > NHL for Win Shares = 0.79 * 0.86 = 0.679

                   SHL > AHL > NHL for Win Shares = 0.99 * 0.87 = 0.861

Using the weighted elements NHLe = (0.679 * 39%) + (0.861 * 61%)

Which gives us NHLe = 0.79 (using more decimals everywhere) for the SHL.

Down the ladder

Then we repeated the process for all the other leagues, as soon as their needed paths to the NHL had been established. Next in line was the Swiss National League, with a first case of small sample size, as Swiss goalies do not leave Switzerland and foreign goalies used to be rare because of the previous cap on import players. There we had to use all links towards Liiga, AHL, NHL and KHL, each with a sample size of 3 to 5 but that’s all we could be doing.

As we went down the ladder, the networks do not necessarily get more complicated, we just needed to connect more leagues together. For many of the smaller leagues, only a handful of paths were necessary, like connecting level 2 league in a country to their level 1 league, or junior league to senior league.

And we arrived at the following ranking.

The first thing that jumps to the eyes is how much higher those NHLe are compared to the players’ NHLe where KHL was at 0.71 and then AHL at 0.54.

Well, it would seems the gap between talent is just narrower for goalies.

Second thing: many leagues are very close to each others. And when you pay attention to European hockey for instance, it should not surprise you. There is not much difference, on average, between a SHL goalie or a DEL goalie. Also those guys are largely playing in their home country and do not travel a lot, so they don’t all regroup in a one or two leagues for instance, basically because there are not enough starting spots available…

One example was given in Switzerland last season, when after raising the number of non-Swiss players allowed from 4 to 6, many teams saw an opportunity to get a foreign goalie, and a lot of Finnish netminders came in. Were they better than the average Swiss goalie? For sure. There suddenly was a new handful of starters matching the top Swiss goalies. But it did not prevent Geneva to win the title with a tandem of two Swiss goalies, changing their starter mid way through the Playoffs…

How relevant are NHLe?

Last but not least, how good are these NHLe compared to actual direct links between two leagues? Meaning, if we compare the NHLe of two leagues, say Poland (NHLe 0.44) and Denmark (NHLe 0.59), I can derive a coefficient for goalies going from Poland to Denmark of 0.44 / 0.59 = 0.75. If I want to compare these coefficients derived from NHLe values with actual coefficients existing between two leagues, using leagues with at least 10 links between them, I get a r2 = 0.87. Pretty sweet.

Like that NL example above, where we had to use small sample size links, ending with a 0.80 NHLe. Historically, two goalies went straight from the NL to the NHL, with a 0.82 NHLe, we we are really close here.

So now we have a tool for further analysis… Like projecting a goalie in a new league with no precedent to based yourself on? Not a problem anymore.