NHL Equivalency and Prospect Projection Models: Analyzing the 2021 Draft Class (Part 4)

How Powerful is This Draft Class?

Photo by EyesWideOpen/GettyImages

The overwhelming consensus regarding the 2021 NHL Entry Draft Class is that a pick in this year’s draft is not as attractive as the same pick in last year’s draft would have been or the same pick in next year’s draft will be.

What isn’t so clear is whether this consensus exists because the draft class itself is weak, or simply because scouts have seen less of this year’s prospects due to the pandemic.

According to Dylan Griffing, a Hockey Scout at Elite Prospects who covers Russian players in particular, the consensus exists mostly due to the former:

However, when Elliotte Friedman discussed the Buffalo Sabres and their views on the draft in 31 Thoughts, he suggested the latter is actually mostly why NHL teams in particular hold this sentiment:

Friedman is probably somewhat inclined to massage the truth here. Even if he believed the consensus among NHL teams was that this draft class is trash, I wouldn’t expect him to just come out and say these kids all suck at hockey. But my expectation doesn’t change that the greater concern voiced by hockey men doesn’t appear to be that these players aren’t so good, but just that they haven’t seen as much of them.

I chose to take a more quantitative, fact-based approach to answering whether the draft class itself is actually weak or just full of lesser known players. The methodology I used to determine the strength of this draft was actually quite simple, and is laid out below:

  1. Treat every player’s probability of becoming an NHLer and/or NHL star as a decimal which represents a number of “projected stars” and “projected NHLers.” For example, a player with a 50% probability of becoming a star and 95% probability of making the NHL represents up 0.5 projected stars and 0.95 projected NHLers.
  2. Include all players who played at least 5 games in one league in their draft year, but use the output from the model which incorporated a player’s draft-1 and draft years if they played both.
  3. Compare the total number of projected stars and projected NHLers in this year’s draft class to those in every other draft class since 2007.

Here are the results:

╔════════════╦═════════════════╦══════════════════╗
Draft Year ║ Projected Stars ║ Projected NHLers
╠════════════╬═════════════════╬══════════════════╣
║ 2007 ║ 12.76 ║ 48.25 ║
║ 2008 ║ 12.93 ║ 50.74 ║
║ 2009 ║ 15.83 ║ 53.91 ║
║ 2010 ║ 16.08 ║ 56.03 ║
║ 2011 ║ 14.27 ║ 55.64 ║
║ 2012 ║ 11.39 ║ 47.52 ║
║ 2013 ║ 16.89 ║ 56.04 ║
║ 2014 ║ 14.39 ║ 52.22 ║
║ 2015 ║ 15.02 ║ 58.88 ║
║ 2016 ║ 15.89 ║ 55.44 ║
║ 2017 ║ 12.47 ║ 51.55 ║
║ 2018 ║ 13.02 ║ 49.67 ║
║ 2019 ║ 13.02 ║ 49.38 ║
║ 2020 ║ 15.76 ║ 51.72 ║
║ 2021 ║ 10.46 ║ 39.21 ║
╚════════════╩═════════════════╩══════════════════╝

The data unequivocally supports Griffing’s position: This is a very bad draft class. It contains fewer projected stars and fewer projected NHLers than any other draft I have run my model on.

The data also bears out that in years for which we can benefit from hindsight, the model did a good job of analyzing draft cohorts based on how they scored prior to their draft date. The 2015 draft class is generally considered the best of this bunch, and it contained the most projected NHLers by a decent margin. By contrast, the 2012 draft class is widely regarded as the worst of this group, and it is the worst pre-2021 draft according to the model.

2021 is a different animal entirely, though; even the horribly underwhelming 2012 draft class looks closer to average than it does to 2021.

I also don’t buy the argument that this class only looks bad because a large percentage of them didn’t play. First and foremost, I included the U18 and U20 World Championships, so prospects like Ethan Del Mastro and Wyatt Johnston who played in only those tournaments are still accounted for here. Second, I spoke again with Dylan Griffing, and he named only four skaters on his board who did not play competitive hockey this year. Here are those four skaters and their respective star/NHLer probabilities according to the Draft-1 Model:

  • Ty Voit: 1.3% star probability, 4.6% NHLer probability.
  • Connor Lockhart: 0.4% star probability, 1.3% NHLer probability.
  • Artyom Grushnikov: 0.1% star probability, 0.8% NHLer probability.
  • Colby Saganiuk: 0.1% star probability, 0.2% NHLer probability.

These players would add a grand total of 0.02 projected stars and 0.07 projected NHLers to this year’s draft. That would make a negligible difference and still leave this year’s draft far behind every other year.

I won’t deny that there is an additional degree of uncertainty attached to the projections made for each player — one which exceeds the uncertainty we see in normal years. This additional uncertainty may skew the “true” probability of success in either direction more so than it would in most years. Perhaps this even further justifies the bearishness towards the draft that we see from NHL executives. But make no mistake: The biggest issue with this draft cohort is just that they’re not good. Scouts don’t need to watch players in order for those players to score, and this cohort is scoring at an unprecedentedly low rate. Teams slated to draft high in 2021 have walked into a relatively unfortunate position.

Making the Most of a Bad Situation

At the conclusion of my last article, I stated that the rate at which a defenseman scores as a prospect is very important, and I greatly recommended against using high draft picks on defensemen who aren’t proficient scorers.

This ties back in to the 2021 draft, and more specifically Owen Power. Power is currently the unanimous number one in the draft according to Bob McKenzie’s TSN’s Mid-Season 2021 NHL Draft Rankings.

While I understand what makes Power’s toolset intriguing to scouts, I’m less enamored by his tools than I am concerned by his unspectacular scoring over the past two seasons (and subsequently low probability of becoming a star). According to this model, there are ten players in this year’s draft class with a higher probability of becoming a star.

The model unequivocally demonstrates that Power is a very good prospect; we’re not talking about Mirco Mueller here. Power is a virtual certainty to make the NHL and be an above replacement level contributor, and his 18% probability of becoming a star is nothing to sneeze at either. But with the option to pick any player available — which, even in this relatively weak draft, includes the option to pick one of two players whose probability of becoming stars are over 3 times Power’s — I believe selecting Power 1st overall is the wrong decision.

I understand it’s not easy to just take these probabilities as the ultimate source of truth. I recommend against using the outputs of anybody’s model in this manner, and I don’t consider my own models an exception to the rule. Although part 3 of this series laid out the strength of these metrics by demonstrating their performance in testing, just saying “William Eklund has a 71% probability of becoming a star and Owen Power has a 18% probability” will understandably not sway the opinion of anybody who already has their heels dug in on drafting Power. It shouldn’t.

A more effective exercise to convey the meaning of these metrics may be to look at some of Owen Power’s statistical comparables: Other defensemen with a star probability between 10% and 30% according to the Draft-1 and Draft Year model:

(I included all players who were drafted before 2016, as I feel it’s still too soon to determine what everybody else will become.)

There are some damn good defensemen on this list. I see a few stars. It’s solid company. But solid company isn’t what you want to hear when you’re talking about a 1st overall pick. And outside of maybe Dougie Hamilton (who I’m not quite as high on as the rest of the analytics community is), I don’t see a single one of them who I would consider a success if I had drafted them 1st overall. I also don’t see a single one who I would even consider taking 1st in a re-draft. Every defenseman who appears on this list comes from a draft where somebody notably better was also drafted.

Using the same method of subjectively looking at statistical comparables can tell us a lot about the top three forwards in the 2021 draft according to the model: Dylan Guenther, William Eklund, and Cole Sillinger. All 3 of them have a star probability between 45% and 75%, a range which produces the following list of comparable players:

There are two Hart Trophy winners on this list.

There are three players who would at least challenge to be selected 1st overall in a re-draft.

Despite the fact that there are 3 more Owen Power comparables than comparables to the top-3 forwards in this draft (22 and, 19 respectively), there are 3 times as many stars in the comparables to the forwards (3 and 9, respectively) according to the star model.

Among those who have spent any time in the NHL, the players comparable to Guenther, Eklund, and Sillinger have produced 2.15 WAR per 82 games; the players comparable to Owen Power have produced 0.75.

Forgive my phrasing, but there is no comparison to be made between these comparables. You can slice and dice it any way you want, but the data clearly bears out that Power is not the optimal pick here, and it’s well outside the margin of error. Factors like his size are already accounted for by the model, and based on how they’ve historically translated to success in prospects, they don’t manage to make up for his unspectacular scoring profile relative to his peers at the top of the draft.

While the model is very effective, it is worth repeating that whatever degree of uncertainty comes with this data is inevitably exaggerated by the pandemic and the effect that it has had on all of these players. (If the model had been built under pandemic circumstances, I doubt it would have performed as well in testing.) And as with every model I make, I wish not to present any of this evidence or any of my arguments as a closing point to a discussion, but rather a starting point.

With regards to Power in particular, his Draft-1 season was particularly impressive, and I’m open to hearing an argument that external circumstances threw off his scoring this season. I’m also open to an argument that scoring doesn’t tell the full story of his contributions at both ends of the ice, and that the contributions that don’t show up on the scoresheet are extraordinary. But it will take a very strong argument to make me disagree with clear conclusions drawn from good data, which is exactly what I have here.