AN ARGUMENT FOR NHLE

NHLe is an equivalency formula used by some in the hockey analytics community.  It’s a method of standardizing scoring across
various major and junior leagues. 
Standardized scoring gives an idea of how players, generally younger
prospects, perform at the NHL level. Some argue its merit as a valuable metric in assessing future performance.  The following provides a framework of how it can be used as a possible drafting qualifier.

The method was developed by Gabe Desjardins a number of years ago.  You
can also find some more information about NHLe over at Matchsticks & Gasoline, our Calgary Flames blogging comrades.    

The equation looks like this …

Here’s an example of how it works…

The translation for the KHL is
0.82, the translation for the NCAA is 0.41, the translation for the WHL is 0.3
and the translation for high school is roughly 0.07.  Let’s assume that four players scores at the
same PPG in four separate leagues, let’s say 60 points in 50 games (PPG of
1.2).  The players’ corresponding NHLe (when we
scale it over 82 games) would be: KHL – 80.7; NCAA – 40.3; WHL – 29.5; HS –
6.9.  In summary, a soon to be drafted 18
year old player putting up those points in the KHL is a rock star
generational-type prospect, a really good NCAA prospect, a decent WHL prospect
and a very low probability high school prospect.

The knock on using NHLe as a
method to project how a player is going to perform in the NHL is that it’s
rarely completely accurate.  The
equivalency is based on the average of how players have performed in the NHL the
year after coming over and the standard deviation is huge
(20-30%).  This is to say that most
players don’t score exactly where their NHLe suggests.

Where NHLe is beneficial,
however, is as an indicator of future potential offensive scoring, in a general sense.  It’s not perfect but we will see in a minute
why NHLe, and especially high NHLes, can be used to help predict future offense.

This is a list of all the 1st round forward draft picks from 2005 to 2010.  The table is sorted by highest NHLe to lowest.  I’ve highlighted in blue all players that have scored at a rate of 0.6 (approximately 50 points per 82 games) to this point in their career.  This is not to say scoring below this is a bust.  There are many forwards that put up 30-40 points a year but are incredibly valuable for what else they do.  But the thought here is that in the 1st round a team’s main goal is to find an offensive threat.  The threshold for this was arbitrarily set at 0.6 PPG over their career.

What we find from just looking at
the data is that very rarely does a team end up with a player who averages less than 50 points a year (very early in their careers) if the prospect has an NHLe north of 35, this seems to be the “can’t
miss” range.  Perhaps the player doesn’t
end up being the best of the draft but putting up 50 points + year after year
in the toughest league in the world is pretty good. If we dive a little bit deeper, looking at some predictive regression modeling, the story gets better.

Here’s a brief crash course on how regressions work …

Regression is a statistical process for estimating the relationship among variables. The adjusted R² gives an indication of what type of impact one variable or group of variables (i.e., the independent variables) is having on another variable or outcome (i.e., the dependent variable).  You want this as close to 1.0 as possible as it indicates a very strong relationship and a strong model. With regards to the independent variables being assessed in the model, you want a p-value as low as possible (e.g., anything lower than 0.1 is good in this case); that indicates the variable is a significant factor and plays a role in the dependent variable you’re looking at. In this case, the dependent variable is the player’s current NHL ppg.  The independent variables are draft year NHLe, position taken in the 1st round and draft year PPG.

What we find is that this model
explains about 24% of a player’s future PPG in the NHL.  It’s not the whole story but these simple
three variables are explaining about a quarter of the story. In this model, NHLe and the draft position are helping to explain future performance in the NHL whereas draft year PPG is not in the least.  The higher a player’s NHLe and the higher they are chosen (often the highest NHLes go first … but not always) the higher their eventual PPG in the NHL should be. That’s the main take-away.

Things like the
player’s ability to drive possession, skating ability, compete level, attitude,
willingness to learn, how they’re developed after being drafted, QoT and QoC would likely make up the rest of this model.
Those figures are much more difficult to track, especially at the
junior/development level, however. One day we’ll get there!

Back to this model. As you look back
over drafts you’ll often find players that end up “surprising” and far exceeding expectations
based on where they were chosen (e.g., Giroux, P. Stastny, M. Savard, T. Fleury,
Ribeiro, etc, etc, etc.) have the draft NHLe to
suggest that they should be a very good player (35+).  They are skipped over for one reason or
another (size, playing in an unknown overseas league, age, attitude, etc.) but surprises are more often than not
players that should have gone higher if their NHLe was one of the main
qualifiers for drafting. 

To illustrate this point, if you reverse the process and look at the highest scorers in the NHL over the past 20 years you find that draft year NHLe still comes out as a significant predictor (R²: 0.16; p-value: 0.002) whereas the significance of draft position goes away. There’s enough late-round surprises in the data to tip the scale the other way. 

Regardless of how you slice
it, junior/developmental league scoring is, in my mind, the single biggest predictor of future offensive
performance. These numbers, just simple
standardized counting principles, can be leveraged by NHL teams to draft better
by taking a significant portion of the guessing out.  Maybe
a team loses out on getting “the best player” here and there but if a team follows these
principles they would likely draft a very, very good player again and again and
again.  

Start to factor in metrics like
age (i.e., younger 35+ NHLes are better than older 35+ NHLes), massive NHLe
jumps from their draft-1 year to their draft year (i.e., players that jump by
10-20 points) and individual primary, ev and total team point totals and the
picture becomes clearer.  All this before ever even seeing a player play.

This year, there’s five players (North American skaters anyways) that fit the “can’t miss” 35+ NHLe mold in the draft class. They are: Sam Reinhart (43.1), Leon Draisaitl (40.4), Sam Bennett (39.3), Nikolaj Ehlers (39.3), Robert Fabbri (36.9). Dal Colle (34.9) is right there as well. Over-ager Louick Marcotte (35.5) also fits the bill but he’s two years older than the rest of the crop.  He could be worth a late-round flier though. I bet we hear a lot about all of these guys in the coming years and they turn into very productive players for their respective teams.