Quantitative 2021 NHL Entry Draft Performance Rankings

Put down the remote control and pick up a spreadsheet!

With the 2021 NHL Entry Draft now officially in the books, it’s time for me to complete my final chapter of content from this draft and use the results of my NHL Equivalency and Prospect Projection Models to analyze how well everybody drafted. If you’re unfamiliar with these models, I suggest you read the 4-part series where I documented how they are built. But if you haven’t read them and you don’t want to, here’s a high-level overview:

  • NHL Equivalency (NHLe) assigns different values to points scored in 124 different leagues based on how players who play in multiple leagues in the same season have historically scored.
  • The Prospect Projection Model is a regression model built using NHLe in addition to age and size to predict the probability that a prospect will become an NHLer and/or an NHL star.

For the sake of simplicity, I chose to rank prospects based solely on the probability that they would become a star. It stands to reason that NHLer probability should also hold some weight, but with the clock ticking and no good idea of how to weight these probabilities, I simply chose to roll with star probability and re-visit this at a later date. I will use the same criteria to rank each NHL team’s draft performance.

Before I move forward with the rankings, I must address the elephant in the room: Montreal selecting Logan Mailloux 31st overall despite Mailloux attempting to renounce his rights prior to the draft. For those unfamiliar with what led Mailloux to renounce his own rights, Habs Eyes on The Prize published this article which outlines it clearly (Trigger Warning: The Article Discusses Sex Crimes).

The consequences and implications from this selection are far worse than any statistical model I’ve put together is capable of measuring. As such, I’ve decided to leave this selection entirely out of my draft performance reviews, just as I left Mailloux entirely out of my final draft rankings. Regardless of what the model says, make no mistake: Montreal made the worst decision of the 2021 NHL Entry Draft.

In addition to Logan Mailloux, I also left all goaltenders, overage players, and players who did not play in 2021 out of my rankings. Although I did not include overage players in the rankings, I still have model outputs for them, so I am still able to analyze those picks and their expected value. Going forward, “picks” will refer only to picks that I am capable of analyzing. A team like the San Jose Sharks may have drafted 9 players, but because one was a goaltender and two did not play this year, I will say they have 6 “picks.” By ignoring all goaltenders and players who missed this season, my methodology stands to underrate teams who used valuable draft picks on good non-analyzable players and overrate teams who used valuable picks on bad non-analyzable players.


I mentioned earlier that I would only use the outputs of the projected star model to analyze each team’s performance. But I can’t just add together the number of projected stars each team drafted and call it a day; that would underrate the performance of teams like Toronto who only had a few picks (which came later in the draft), and overrate the performance of teams like Columbus who had a ton of high selections high in the draft.

In order to properly compare the performance of teams who entered with a different amount of draft pick capital, the opportunity cost of the picks they made must also be taken into consideration. I calculated the opportunity cost of these picks by adding the star probability of the top player available on the board at each team’s selection, and then comparing the sum of those values to the sum of the star probability for all the players they took.

As a quick example, say San Jose made three picks with star probabilities of 50%, 20%, and 30%. The value of these players could be respectably interpreted as 0.5, 0.2, and 0.3 “expected stars” and the total “expected stars” from San Jose’s draft would be 1.

Now say that at the respective time they made those three picks, the best player on the board (by the projected star model) held star probabilities of 80%, 70%, and 50%. This would mean that they had 2 total expected stars available to them, and drafted 0.5 expected stars for every 1 available expected star. The number of expected stars drafted divided by the number of expected stars available is the measurement I chose to rank everybody’s draft performance.

Here is how each team performed by this measure:

Team ║ xStars Available ║ xStars Drafted ║ Drafted/Available
║ S.J ║ 1.15 ║ 0.75 ║ 0.65 ║
║ ANA ║ 1.70 ║ 0.82 ║ 0.48 ║
║ CGY ║ 1.13 ║ 0.53 ║ 0.47 ║
║ CBJ ║ 2.18 ║ 1.01 ║ 0.46 ║
║ L.A ║ 1.35 ║ 0.45 ║ 0.34 ║
║ TOR ║ 0.41 ║ 0.14 ║ 0.34 ║
║ ARI ║ 1.93 ║ 0.65 ║ 0.34 ║
║ WPG ║ 0.81 ║ 0.23 ║ 0.29 ║
║ DAL ║ 1.16 ║ 0.30 ║ 0.26 ║
║ NSH ║ 0.99 ║ 0.16 ║ 0.17 ║
║ T.B ║ 0.70 ║ 0.10 ║ 0.15 ║
║ N.J ║ 1.38 ║ 0.21 ║ 0.15 ║
║ EDM ║ 0.67 ║ 0.09 ║ 0.14 ║
║ BUF ║ 2.20 ║ 0.30 ║ 0.14 ║
║ FLA ║ 0.81 ║ 0.10 ║ 0.13 ║
║ SEA ║ 1.38 ║ 0.17 ║ 0.12 ║
║ MTL ║ 0.99 ║ 0.11 ║ 0.11 ║
║ STL ║ 0.58 ║ 0.06 ║ 0.10 ║
║ COL ║ 0.81 ║ 0.07 ║ 0.09 ║
║ PIT ║ 0.67 ║ 0.05 ║ 0.08 ║
║ NYR ║ 1.08 ║ 0.08 ║ 0.07 ║
║ MIN ║ 0.90 ║ 0.07 ║ 0.07 ║
║ CHI ║ 1.16 ║ 0.09 ║ 0.07 ║
║ WSH ║ 0.58 ║ 0.04 ║ 0.07 ║
║ OTT ║ 1.12 ║ 0.07 ║ 0.06 ║
║ DET ║ 1.38 ║ 0.09 ║ 0.06 ║
║ BOS ║ 0.67 ║ 0.04 ║ 0.06 ║
║ CAR ║ 1.48 ║ 0.09 ║ 0.06 ║
║ NYI ║ 0.67 ║ 0.03 ║ 0.05 ║
║ PHI ║ 0.67 ║ 0.03 ║ 0.05 ║
║ VAN ║ 0.58 ║ 0.02 ║ 0.04 ║
║ VGK ║ 0.81 ║ 0.04 ║ 0.04 ║

At a first glance, this looks about right to me. But further examination of each team’s performance is required to see why each team ranked where they did.

1st — San Jose Sharks

At the age of 18, William Eklund scored 23 points in 40 games in the SHL, one of the world’s top men’s hockey leagues. This extremely impressive, rare feat makes Eklund the best player in the draft according to the model.

By simply using the 7th overall selection on Eklund, San Jose ensured they’d have the best draft by this measure; the rest of their selections which they used on analyzable picks were low enough in maximum available value that even if they used them all on janitors with star probabilities of 0%, their draft performance would still be the best by this measure. It may seem strange to give a team this much credit just for making the obvious pick early, but making a mistake there is where teams stand to lose the most value.

2nd — Anaheim Ducks

It’s not often that the a team’s 1st round pick has a lower probability of becoming a star than their 2nd and 3rd round pick, but that is the case with Anaheim here. Mason McTavish, who scored 9 goals in only 13 games playing against grown men in the Swiss League, is a great prospect in his own right; the fact that Anaheim drafted two players with higher star probability is more a testament to their strong performance than an indictment of that pick itself.

It truly is a wonder that Sasha Pastujov dropped all the way to 66th. Even putting aside all the quirks of a statistical model, he is a 6', 183 lb forward whose consolidated draft ranking was 32nd, and he scored well over a point per game in the USHL, USDP, and U-18 WJCs.

3rd — Calgary Flames

Matthew Coronato nearly cracked a goal per game in the USHL’s regular season, scoring 48 in 51 games, and then eclipsed that mark in the playoffs with 9 goals in 8 games. Players who score like Coronato did have historically turned out very well, and it’s not hard to see why a statistical model gives so much credit to Calgary for putting aside size concerns and going with tangible production.

Aside from Coronato, Calgary also made a few nice value bets in Cole Huckins and Cameron Whynot. It is a bit strange that Huckins, a young, 6'3 forward who scored just under a point per game in the QMJHL, dropped all the way to the 3rd round.

4th — Columbus Blue Jackets

You may be starting to notice a common theme: Every team who grades out very well by this measure drafted the BPA at least once. Columbus went with Cole Sillinger, the 2nd best player in the draft.

Sillinger scored over a point per game in the WHL in his pre-draft year and followed it up by scoring over a point-and-a-half per game in the USHL this year. Between his solid scoring rates and the upward trend in them, his relatively young age, and the fact that he’s already over 200 lbs, this player has absolutely no red flags which carry over to the spreadsheet.

Kent Johnson was also a fine pick with the 5th overall selection, and Corson Ceuelemans was a nice value bet at 25.

5th — Los Angeles Kings

Call it a coincidence, but public prospect models have loved nearly every pick the Los Angeles Kings have drafted ever since they hired Rob Vollman, NHL equivalency guru, as a Senior Hockey Analyst. If there’s anybody capable of quantifying just how impressive Clarke’s 15 points in 26 games against grown men in Slovakia were, it’s him.

Not only did LA select the best defenseman in this year’s draft in Clarke, but they also made 3 other picks with no less than a 1-in-100 chance at becoming stars. A very nice bit of business.

6th — Toronto Maple Leafs

Toronto only made one pick that I’m capable of analyzing, but it was a damn good one in Matthew Knies, who scored just over a point-per-game in the USHL in his pre-draft year — a feat typically observed from future 1st round picks.

While it is concerning that Knies took a step back in his draft year, scoring just under a point per game, the fact remains that he is a 6'3 forward whose production in both years was solid at worst. At 57th, he’s a very good value bet. Quality over quantity.

7th — Arizona Coyotes

Dylan Guenther is the crown jewel from Arizona’s draft, which makes sense: He scored 12 goals and 12 assists in 12 WHL games this year. This feat should be taken with a grain of salt, and would mean a lot more if he’d kept this pace over a traditional schedule of 50 games, but it’s still very impressive.

Say what you will about Arizona’s picks outside of Guenther, but you can’t deny they were bold. Using the 37th overall pick on overager Josh Doan, who just so happens to be the son of Shane Doan, Arizona’s current Chief Hockey Development Officer, was bold. Using the 60th overall pick on overager Janis Jerome Moser, who had already been passed up on in not one, not two, but three consecutive drafts, was also bold.

I believe one of these bold moves was actually quite shrewd. Moser scored 30 points in 48 games against grown men in the Swiss League this year, the kind of production you’d expect out of a top-5 pick if they were a forward in their draft year. Moser is, of course, three years removed from his draft year, but he’s also a defenseman. The fact that he’s 21 years old means there’s a lot less room for development, but his accomplishments at the age of 21 also trump the accomplishments that the average 60th overall pick will reach at any age. It’s certainly a strange decision, but the data supports it.

8th — Winnipeg Jets

Chaz Lucius played at least a dozen games in both the USHL and USDP this season and scored at least a goal per game in both. As with many players in this year’s draft, he accomplished these feats in a small sample, and he also didn’t record many assists, which means his overall NHLe isn’t that high. But at 6', with a relatively young birthday, this is a solid value bet with the 18th overall pick.

Nikita Chibrikov, who captained Russia at the U-18 WJCs and scored 13 points in 7 games, was also a solid pick with some legitimate upside made at a point in the draft where many teams already give up on drafting players with high ceilings.

9th — Dallas Stars

It says something about the rest of Dallas’ picks that they selected Wyatt Johnston 23rd overall and worked their way up to a top-10 draft performance in spite of it. Johnston is an interesting case: He only played in 7 games this year, all at the U-18 WJCs, and scored only 4 points in those games. But even prior to this wacky season, the 30 points in 53 OHL games his pre-draft year don’t necessarily scream first round pick either. He’s the kind of player that I’d be open to putting aside the model outputs for and taking a chance on, but I’d need to be pretty confident in what else I see in him in order to use a first round selection on him.

On the flip side, Logan Stankoven is also the kind of player I would be open to putting aside the model outputs for and passing on early in the draft. He’s a small forward whose impressive scoring rates came in only 13 games this season. But he did score 48 points in 59 games in the WHL in his pre-draft year, so he’s got a track record of scoring better than Johnston.

Where Dallas really gets a leg up on the competition, though, is with the selections of Ayrton Martino and Conner Roulette in the later rounds. Had they picked somebody with a better statistical profile than Wyatt Johnston with their first selection, they’d be very close to the top. As it stands, they still did a nice job.

10th — Nashville Predators

Nashville is another team whose player with the best statistical profile was not the first they selected. Fyodor Svechkov had a solid draft year performance, but his uninspiring pre-draft year holds him back a bit in the final model outputs. Zachary L’Heureux, on the other hand, has been a bit more consistent from year to year, despite playing in a league where I may under value points a bit.

Ryan Ufko, a defenseman who scored 39 points in 53 USHL games, was also a very nice value pick at 115.

11th — Tampa Bay Lightning

The Tampa Bay Lightning paid a very small price to win the Stanley Cup: They didn’t get to draft until all the good players were already gone. They still managed to get some good value picks, though, including Dylan Duke who scored just under a point per game in the USDP.

Daniil Pylenkov doesn’t have a player card, but he’s a 20, soon to be 21 year-old defenseman who has already been passed on twice, and he scored 19 points in 54 KHL games this year. A 20 year old defenseman who has proven they can contribute some offense in the 2nd best hockey league is not a bad bet at 196.

12th — New Jersey Devils

Would New Jersey have selected Luke Hughes 4th overall if he were not the brother of Jack, their star center selected 1st overall in 2019? I don’t know for sure, but the selection leaves a bit of a bad taste in my mouth. I feel that when businesses decide to prioritize bringing in family members or friends of current employees, they inevitably risk compromising the efficiency of their business operations. Put it this way: What are the odds that the most qualified available person for the job just happens to be related to somebody who is already in the organization?

Luke Hughes’ statistical profile is excellent in and of itself, but it’s a little underwhelming given the high selection and other players available on the board. Again, I don’t know exactly what the Devils would have done otherwise, but I’m far from convinced they would have made this selection if he were not Jack’s brother.

While the Hughes selection was not so poor that a team would necessarily need to “make up” for it, the Chase Stillman selection was. Stillman’s 16 points in 8 games in his draft year sound impressive until you realize he did it in Denmark U20, a horrible hockey league. While it stands to reason that you can only score so many points in any given league because of the limitations of ice time, I’d still expect a forward who is a 1st round talent to score more than 2 points per game in such an awful league.

New Jersey did make up for Stillman with their other selections, though. Samu Salminen, who scored a goal per game in the U-18 WJCs, was a nice pick in the early 3rd round. Topias Vilen, a defenseman who scored 8 points in 35 games against grown men in Liiga, was a nice pick in the 5th.

13th — Edmonton Oilers

Luca Munzenberger may be the most puzzling selection any team made in this year’s draft. While it’s likely that scoring underrates the talent and potential of a defensive defenseman who played in a terrible league like the DNL U20 and did not score at all in 5 games in the U-18 WJCs, it’s not as though this is a player who invoked wars between scouts and spreadsheets; the only scouting service who even ranked him was McKeen’s Hockey, who ranked him 214th. It’s one thing to see something in a weak offensive player and draft him late, but to use a pick in the top 3 rounds on a player who may not have been drafted at all is insane.

The selection of Munzenberger was ultimately not of much consequence, though. Xavier Bourgault was a solid pick at 22; an effective scorer in the QMJHL two years in a row.

Matvei Petrov was also a nice selection at 180th. The Oilers didn’t do terrible as a whole, but the selection of Munzenberger is objectively hilarious.

14th — Buffalo Sabres

The selection of Owen Power felt inevitable, even if he was not the player that I would have picked. Power is a great prospect in his own right, but I just felt there were players on the board with significantly higher upside.

Moving on from Power, Isak Rosen is actually Buffalo’s pick which looks by far the worst according to my model. I want to take a closer look at him.

Rosen played at least 5 games in 3 different leagues (which I have modeled) this season: SHL, J20 Nationell, and U-18 WJCs. It was a tale of three different seasons.

  • In the SHL, his 1 point in 22 games (which make up the bulk of his sample size) have mostly been blamed on low ice time. While I don’t believe this entirely excuses his performance (and also believe those invoking this excuse are not properly considering that coaches generally play their best players), I’m willing to accept that external context really harmed his numbers here. Still, it’s not great.
  • In the U-18 WJCs, he scored a goal per game, which is excellent. He scored 9 points total in 7 games, which isn’t elite, but it still more or less justifies the selection.
  • In J20 Nationell, he scored 12 points in 12 games.

Put aside the SHL and the U-18 WJCs for a moment, because I want to focus on Rosen’s scoring in J20 Nationell. If a forward scored 25 points in 38 games in the OHL in their pre-draft year, and then scored 12 points in 12 games in the OHL in their draft year, would that justify a top-15 selection? Probably not. Now consider that Rosen did those things in J20 Nationell, which is a significantly weaker league than the OHL. (The OHL’s league equivalency of 0.144 is 1.58 times higher than J20 Nationell’s equivalency of 0.091.)

Statistically speaking, Rosen is probably not as bad as the model suggests. In a normal year without a pandemic, he may not have played in the SHL, and instead played in a different league where he scored more. But his scoring does not justify this very high selection. At all.

All that being said, Buffalo redeemed themselves from a statistical perspective when they selected Oliver Nadeau 97th. Nadeau scored at a significantly higher rate in the QMJHL this year than Rosen did in J20 Nationell despite the QMJHL being a significantly better league than J20 Nationell, which serves to amplify the head-scratching nature of the Rosen selection.

15th — Florida Panthers

Even on a pick-by-pick basis, Florida had a pretty unspectacular draft. No awful picks, no great ones. Matthew Samoskevich basically personifies their draft.

He was from a bad pick where he went, and he has some legitimate upside, but there were also better players on the board.

16th — Seattle Kraken

Even Doug Wilson Jr., San Jose’s Director of Scouting, who drafted William Eklund, said they ranked Matthew Beniers above Eklund on their draft board. If Matthew Beniers really is as good as scouts unanimously seem to agree he is, then Seattle did a lot better than my model suggests. Because my model suggests Beniers is a great prospect in his own right, but not quite worthy of this selection.

Seattle did okay on the Beniers pick and okay on the rest of their picks as well. Nobody they picked had extremely low upside — even Justin Janicke, who they selected 195th overall, is an intriguing forward who scored at a solid rate in the USHL and USDP despite being one of the younger players in the draft class.

17th — Montreal Canadiens

With back-to-back picks at the end of the 2nd round, you’d liked to have seen Montreal go with somebody with a little more upside than Riley Kidney or Oliver Kapanen. They more or less made up for this, though, by selecting Joshua Roy with the 150th overall pick.

Xavier Simoneau is another intriguing player. He is an overager who had already been skipped twice despite scoring over a point-per-game in the QMJHL in his draft year. At 5'7, 174 lbs he’s quite small to begin with, and at 20 years old, he’s not likely to get much bigger. It will be fun to watch whether he’s able to carve out a niche for himself in the NHL despite his size.

18th — St. Louis Blues

In a model that is fairly low on the QMJHL, it’s not hard to see why a forward who has scored just about a point per game in that league in each of the past two years does not grade out especially high.

Bolduc is not an awful selection at 17, but he is a slightly disappointing selection given the talent that was still available on the board. St. Louis made up for some of it by selecting Tyson Galloway late in the draft, but this still isn’t a great draft performance according to the statistics.

19th — Colorado Avalanche

Oskar Olausson’s scoring in his pre-draft year (1 P/GP in J18 Elit and 0.76 P/GP in J20 Nationell) was not the type you typically see from a forward who is selected in the first round, so his but he showed enough progress in his 2nd year to mostly justify the selection, posting solid production across 3 different Swedish leagues which added up to an NHLe of 13.

The strategy of prioritizing skating and puck skills in defensemen regardless of their stature has mostly worked for Colorado so far, so it’s no surprise they went back to the well with Sean Behrens. What is surprising is that nobody else followed Colorado’s lead and snagged Behrens before 61st.

With that being said, making a sentimental selection on Taylor Makar — Cale Makar’s brother — harms Colorado’s overall performance a bit, even if he was the 220th overall selection. There were significantly better players available with that pick than a 20 year old forward who barely broke a point per game in the AJHL.

20th — Pittsburgh Penguins

Pittsburgh made a nice pick with Tristan Broz, who scored just under a point per game in the USHL, at 58th. His statistical profile was that of a player who typically goes a bit higher in the 2nd round or sneaks into the first.

Outside of Broz, though, this was an unremarkable draft. Pittsburgh selected four other players, and only one of them even rounds up to a star probability of 1%. For a team that needs to start re-stocking the cupboards soon, they probably could have swung a little harder for the fences. (Although Isaac Beliveau’s scoring in his pre-draft year suggests he may be a bit better than this projection gives him credit for being.)

21st — New York Rangers

Brennan Othmann was everybody’s first introduction to the fact that some players with very low star probabilities will go very high in the first round. As of this moment, the Tweet where I announced this selection and attached Othmann’s player card has over 100 Quote Tweets.

Rangers fans were, of course, very unhappy with the model output, but it’s not hard to understand why Othmann’s statistical profile looks this way. He scored 33 points in 55 OHL games in his pre-draft year, which is notably less than we typically observe from a forward who goes on to be selected in the first half of the first round. And in his draft year, he went over to the Swiss B League, where my work has found that a point is worth only 0.176 NHL points, and scored under half a point per game. At the end of the day, he is not a great scorer, and at 6' 174 lbs with a January 2003 birthday, nothing else about him jumps off the spreadsheet either. I’m sure the Rangers see something in him that I don’t, but I can’t fault a model for being low on him.

New York did a nice job selecting Ryder Korczak 75th overall, but it wasn’t quite enough to redeem the Othmann selection.

22nd — Minnesota Wild

The consensus from draft people on Twitter seems to be that Jesper Wallstedt is an incredible prospect and a steal with the 20th overall selection. If you’re of that opinion, that’s awesome, and you may be very well justified in saying Minnesota had an excellent draft. I’m personally not qualified to have any sort of opinion on Wallstedt or the selection, as I’ve neither seen him play nor worked on goalie statistics.

The chunk of Minnesota’s draft which I can statistically analyze was a bit underwhelming, though. Lambos had a bit of a weird year, and might be better than the model suggests, but it’s still probably fair to say that Minnesota started the draft by picking two defensemen with relatively low upside. Caedan Bankier, though, looks like a nice pick.

Bankier scored 23 points in 22 WHL games this year. It’s a small sample, and he scored much worse in his pre-draft year, but given that he’s a normal sized forward, I probably wouldn’t have let him slip this far.

23rd — Chicago Blackhawks

A common theme of this year’s draft was teams drafting relatives of members of their organization; Arizona drafted Shane Doan’s son, New Jersey drafted Jack Hughes’ brother, etc.

The data did not support any of these decisions. But it actually does support the selection of Colton Dach, Kirby Dach’s brother, with the 62nd overall pick. Dach is a 6'4 forward who scored a point per game in the WHL this year; it’s a bit puzzling that he fell this far.

Of course, Chicago needed to do something good here in order to make up for selecting Nolan Allan — a defenseman who scored 2 points in 16 WHL games this year — with their first round pick. And with 5 other very low upside selections, their draft as a whole still looks rather poor.

24th — Washington Capitals

As a defenseman who scored 32 points in 47 USHL games this year, Brent Johnson represents some legitimate upside, and a solid pick at 80.

Outside of Johnson, this is a pretty uninspiring draft, but Haakon Hanelt is an interesting player I’d like to focus on. The only data point used is his DEL play this year, where he scored 1 point in 22 games. Maybe he’s another young player who got forced into a checking role against grown men, and maybe he did better than the stats suggest?

25th — Ottawa Senators

The Tyler Boucher selection actually looks better according to my data, which ranked him 30th, than it does according to the vast majority of scouts who had him outside of the first round entirely. But it still does not look great.

At the end of the day, it’s going to be hard for a statistical model to justify selecting a forward 10th overall when they score only one point per game in the USDP in their draft year.

It really feels like Ottawa isn’t even trying to extract maximum value out of their assets here. It seems very unlikely that Boucher would’ve been gone by 20, so if you really do feel Boucher is the best player available at 10, why not trade back with a team in the teens?

The same can be asked of the decision to select defenseman Benjamin Roger, who was ranked no higher than a late 3rd round pick, with the 49th overall pick. I do not have data for Benjamin Roger, as he did not play this year, but given the fact that he scored only 8 points in 35 OHL games in his pre-draft year, I highly doubt it would look much better; I suspect it would actually look worse.

26th — Detroit Red Wings

I understand that Simon Edvinsson does some things that scouts love. But he played in four different leagues and scored at an impressive rate in none of them. 6 points in 14 games in the J20 Nationell — a league which, as I laid out earlier, is significantly worse than the OHL — is especially unimpressive. At the end of the day, defensemen who score at unimpressive rates as prospects seldom become legitimate stars in the NHL.

Edvinsson is not a bad prospect, but selecting him with William Eklund and many other high-upside players on the board was an awful pick.

Detroit recouped some value with Red Savage in the 4th round, but not enough to make up for the Edvinsson selection.

27th — Boston Bruins

Boston’s poor performance here really comes down to the decision to select Fabian Lysell 21st overall. And Lysell’s very low projection comes down to the fact that he scored only 3 points in 26 games in the SHL.

I’m open to the argument that Lysell’s SHL scoring rates were artificially deflated by low ice time. At the same time, you expect true high-end prospects to force coaches to play them more, and Lysell’s scoring in Swedish Junior Leagues in both his pre-draft year and draft year still don’t really justify the selection.

Lysell did score a respectable 9 points in 7 games in the U-18 WJCs, which would more or less justify this selection. Again, I’m open to the idea that the model underrates him due to low ice time in the SHL. But at the same time, it’s still fair to say his statistical profile thus far in the aggregate does not justify this selection, and Boston did not draft anybody special enough with their other picks to make up for it.

28th — Carolina Hurricanes

After all the timeouts and the trades back, Carolina wound up with a bunch of players who don’t statistically grade out very well according to my models. Scott Morrow played in the USHS-Prep League, which has a hilariously low equivalency of 0.028, and makes it pretty hard for a player to post a high NHLe; it’s reasonable to ask how many more points Morrow would have scored in 30 games than the 48 he already scored if he were a legitimately elite offensive defenseman prospect. But if that’s the best justification for a player’s low NHLe, I’d probably pass on them in the top-40.

On the flip side, a defenseman like Aidan Hreschuk, who scored at a fairly high rate in the USHL and the USDP, is the kind I’d look at in the late 3rd round.

It’s really hard to get a read on Carolina’s draft performance because they selected so many players (including 4 which I am incapable of analyzing), and the number of timeouts, trade backs, and selections they made give off the impression of a genius organization making some moves that just go beyond my understanding. It’s totally possible that there’s something I’m missing, but on paper, it looks like they just kinda galaxy brained it.

29th — New York Islanders

The New York Islanders didn’t have a ton of picks, and they chose to use their highest on somebody whose reputation was that of a player that point totals did not tell the full story for.

Aatu Raty scored at a very impressive rate two years before his draft year, has supposedly done a nice job of driving play in Liiga, and is beloved by scouts, but his NHLe was very poor this year.

Only 12 forwards with an NHLe of less than 7 in their draft year have ever become stars in the NHL. At the end of the day, the Islanders went against the data with this one, and didn’t do enough outside of it to change their overall rankings. More power to Lou Lamoriello, Barry Trotz, and company if they can turn this player into a star.

30th — Philadelphia Flyers

Philadelphia didn’t walk into the draft with much draft pick capital, but also didn’t do much with what they had. Samu Tuomaala’s 11 points in 7 games in the U-18 WJCs is very impressive, but in the much bigger sample of 30 games in U20 SM — Sarja — a weak league where a point is worth 0.083 NHL points — Tuomaala only scored 31 points. He also went scoreless in 5 games in Liiga. At the end of the day, you’d like to see a lot more against such weak competition from a forward you draft in the first half of the 2nd round.

Flyers fans can take some reprieve in the fact that their draft performance grades out poorly, though: At least they probably wouldn’t have drafted anybody good if they kept the 14th overall pick they traded for Rasmus Ristolainen!

31st — Vancouver Canucks

Much like the Flyers, Vancouver traded their 1st round pick hours before the draft in exchange for an overrated defenseman, walked into the draft with low draft pick capital, and then made little to nothing of it. Their first selection, Danila Klimovich, is interesting.

Klimovich’s scoring in the two Belarussian leagues was quite poor, but he managed 6 goals in 5 games in the U-18 WJCs. It also stands to reason that because Klimovich played in such a weak league like Belarus-Vyssha with a league equivalency of 0.052, perhaps even an elite player in his shoes would not have managed too many more than the 52 points in 37 games that he did. This is the type of player absolutely worth taking a flyer on at some point in the draft, but with the 41st overall pick, I’d be very uncomfortable with it.

Much like Flyers fans, Canucks fans can take reprieve in the possibility that if they hadn’t traded their high 1st round pick in an awful deal, they might’ve just used it poorly anyway.

32nd — Vegas Golden Knights

Neither Zach Dean nor Daniil Chayka are bad prospects, but neither hold the type of upside you expect from a player drafted in the 30s. Chayka in particular had an interesting year, bouncing between the KHL, MHL, VHL, and U-18 WJCs, and scoring well in none of them. The 34 points he scored in 56 WHL games in his pre-draft year were actually more impressive than this year’s.

The downward trend is concerning, and by selecting two players with relatively bleak outlooks in Dean and Chayka, and then following them up with two players who don’t even round up to a 1% chance at making the NHL, Vegas put together the worst performance in this draft according to the model. Subjectively, I’m not quite as harsh on them, but I understand why the model is.

Closing Thoughts

There is more to a hockey player than their age, size, and scoring rates; points are a flawed measure in and of themselves, and they are especially flawed this season where the sample of games played is even smaller than usual and the quality of each league is prone to more variance than usual. I have no doubt that NHL teams have access to more information than I do, and that they used more information to form their decisions than I did.

On the other side of the fence, though, is the fact that my limited information all has clearly defined value. Points in different leagues are weighted in the manner most optimal for predicting out-of-sample scoring. League-adjusted scoring rates in the pre-draft year and draft year, along with height, weight, and age are all weighted in the manner most optimal for predicting out-of-sample career outcomes within my Wins Above Replacement model. Everything in my Wins Above Replacement Model is weighted in the manner most optimal for describing past success at the NHL level.

The bottom line is that players who score at higher league-adjusted rates become significantly better players at the NHL level. Can the same be said of players with firm handshakes who say the right things in their interviews? I don’t know. But I know that’s something that NHL teams place weight on, and something that I don’t know or care about.

I’m excited to re-visit this piece in 5–10 years. While I don’t know for sure whether I’ll be mostly right or mostly wrong about what I said here, I do know I’ll learn a lot.

Data Scientist https://www.linkedin.com/in/patrick-b-647077b4/