2019 NHL Draft: By The Numbers

A few months ago on the day of the 2019 NHL draft, I revealed my consolidated draft list compiled from 60 different sources. The reception was far beyond what I expected for a quick draft article — being able to quickly figure out whether a player was a steal or reach on draft day proved to be pretty popular.

It’s since led me down a rabbit hole of scavenging rankings — you can already find the consolidated rankings for the 2020 draft here.

This article will be broken up into two sections: a breakdown of which lists performed the best, and then some interesting stats and info from the 2019 draft.

The Best of the Best

First, let’s look at which sources provided the most accurate draft lists. I judged performance based on two metrics:

• Drafted accuracy: The percentage of the ranked players who were drafted within that range. For example, if a source ranked its top 100 players, and 85 of them ended up being drafted in the top 100, that list would have an accuracy rating of 85%.
• Positional accuracy: The average difference between where the players were ranked and where they were drafted. To avoid penalizing lists that are on the longer side — the difference between 185 and 195 is much less important than the difference between 3 and 13 as an example — I converted every ranking into a draft value, based on research from Michael Schuckers. This isn’t a perfect method since every draft class varies in quality, but taking the average difference between draft values is a reasonable way of levelling the playing field among the lists./

After creating these two metrics, I generated an overall score by combining the percentiles of the metrics for every list.

But before I reveal the best lists, an asterisk is warranted. These metrics are designed to see which rankings fell best in line with what happened on draft day, which isn’t the goal of most draft rankings.

The main purpose of most draft rankings is to give an opinion about which players the source thinks will have the best success in the NHL, instead of trying to play a guessing game in the heads of NHL scouting staffs. There are some exceptions, of course, but the following list shows which rankings came closest to what happened on draft day.

Without further ado, here are the top performers:

2019 NHL Draft - Most Accurate Lists

As expected, Bob McKenzie’s list is tops among the 60 sources in terms of how close it was to what happened on draft day. There’s an obvious reason for why that might be — McKenzie’s list is already a compilation of lists from NHL scouts. His survey of 10 NHL scouts has become widely known as the go-to source for accurate draft information, and this test confirms that.

Last Word on Sports, Derek Neumeier and Canadiens Prospects round out the top five. Whether these sources are reliable at producing accurate lists would require more than one year of data. Although based on 2019 alone, they were all pretty accurate.

Check out the full list of 60 sources to see how each one performed.

One notable list is the consolidated ranking, which performed better than every individual list except Bob McKenzie’s. It’s worth noting that the consolidated list is generated somewhat arbitrarily, as it only includes prospects ranked by 10+ sources, or 155 prospects total, in an effort to strike a balance between having a decent sample size and generating a list with good length. It hit on 138 of 155 players (89%, 2nd best), with an average score error of 49.8 (4th best). For context, a margin of 49.8 roughly equates to the difference between picks #31 and #41.

This was surprising to me, as the public consensus is usually considered to be different from that of most NHL front offices. There’s still a big difference based on the gap — I couldn’t predict things like Moritz Seider being taken at #6 or Arthur Kaliyev falling to #32. But by combining each individual ranking, it eliminated a lot of the noise and ended up performing very well.

Another neat feature of using draft value to create a consolidated list is the ability to find an expected range for every prospect. I go through the process of calculating expected range in this article, with the result generating two numbers, a high and low pick between which the prospect is expected to be taken.

The expected range performed well, with 60 total players being picked within their range (28%). Although with limited rankings for the later rounds, it performed much better in the earlier stages, hitting on 16 players in the first round (52%) and 46 players in the first three rounds (49%). The early second round seemed to be especially rambunctious this year with only two players taken within their expected range between picks #27-42. Many of those picks were widely considered off-the-board, with teams either picking players that fell lower than expected, or picking their guy when they had the chance.

In total, consolidated rankings are a valuable resource not only for determining the public consensus, they also do a good job at predicting which players get taken where on draft day.

To improve draft day predictability, a number of improvements can be made. Constructing a model incorporating elements like player performance, size/weight, and the NHL Central Scouting rankings would likely yield a significantly better performance. But in the context of consolidated rankings, a Kemeny-Young algorithm could provide a more accurate list for the public consensus since it would account for every list containing a slightly different set of players.

It wouldn’t be a full wrap-up without some cool stats and charts! Below are some interesting bits of information.

Wrap-Up Stats

Biggest Steals

Peyton Krebs (VGK) takes the crown as the draft’s biggest steal, taken with pick #17 after being expected to go between 7th and 12th.

Biggest Reaches

It’s no surprise Moritz Seider was the biggest reach of the draft, as the Red Wings took the defenceman with the 6th overall pick when he was expected to go between 13th and 27th.

Total Draft Value

This chart shows which teams accumulated the most overall value at the draft based on the consensus rankings. The Kings came away with the strongest group taking four players who could’ve potentially been a first round talent (Turcotte, Kaliyev, Björnfot, Fagemo).

Draft Value vs. Pick Value

Pre-draft value is determined by the total amount of expected draft value each team had entering the draft. The teams with top picks (New Jersey, NYR, Colorado) are expectedly further to the right. Teams above the diagonal line exceeded expectations while teams below accumulated less value than what they’d be expected to with their given picks.

But How Did The Sens Do?

And finally, because this is a Sens blog, here’s a pick-by-pick breakdown of each player they took.

Reaching on the first four picks of the draft is rarely a move that works out. Whether the Sens’ scouting team had some special vision and saw something in these players that the 60 scouts did not remains to be seen. But based on how the public consensus views their draft, there was plenty of room for improvement.

Looking forward, the 2020 draft is shaping up to be one of the deepest drafts in recent history. The Sens are also loaded with draft picks, having their own 1st round pick this time around, plus that of the Sharks’, plus three picks in the second round.

If you want to get an early idea of what the 2020 draft class looks like, you can check out the consolidated rankings here, or read my profiles of the top 20 prospects here (#1-10) and here (#11-20).