2018-19 DRAFT MODEL UPDATE PART I: DMEN

For the past three seasons, I have utilized a draft model to create my rankings for the NHL Entry Draft. The model consists of using historical rates of return for players based on draft position and statistical measurements. The intent of the model is to find where scouts and stats collide to get the best of both worlds. However, the numbers I have been utilizing in the model have not been updated in the past three years so it was due for an update. Therefore, this summer I spent some hours updating the data to improve the accuracy of my draft rankings. In this post, we will be looking at the results of my work in regards to first year draft eligible defensemen.

DRAFT VALUE FOR D-MEN BY DRAFT POSITION

To determine the draft value for defensemen by draft position, I utilize the career average time on ice (ATOI) as my measurement. While there are many different ways to measure the quality of a defenseman, time on ice is a classic way of rating their value. On the whole, the best defenseman are also the most trusted defenseman and therefore end up playing the most minutes for their team.

The first step is to group the defensemen by finding where there are similar rates of return based on ATOI. For example, first year eligible defensemen selected 10-19 overall, between 1998 and 2011, have a similar rate of return. If you take a look at the scatter plot of the data it has a fairly flat trend line and a R² of 0.0006 meaning that for this sample, draft position explains only 0.06% of that player’s ATOI in the NHL. In terms of ATOI, there is basically no difference between a defensemen selected 10th overall and 19th overall.

Using this method, I divided the 217 draft positions into 10 groupings. Here are the results for games played for these groups entering the 2018-19 season:

At the top end of the draft, you are much more likely to get a long-term NHL player that plays multiple seasons in the NHL. For example, if we chart the percentage of players to play 500 plus games in the NHL it decreases at an exponential rate between group 1 (1-2 overall) and group 10 (163-217 overall).

Meanwhile, if we chart the percentage of players to play less than 100 games in the NHL, it increases steadily from 0% for players taken in group 1 (1-2 overall) to 90.5% for players taken in group 10 (163-217 overall).

Here are the results for ATOI for the same groupings:

Similar to games played, ATOI decreases at an exponential rate between group 1 (1-2 overall) and group 10 (163-217 overall) for defensemen that average 20:00 minutes per game or more and increases fairly steadily for defensemen that average less than 10:00 minutes per game.

DRAFT VALUE FOR D-MEN BY ADJUSTED P/G

In determining the draft value for defenseman by a statistical measurement, I utilize an adjusted point per game measurement. The reason for utilizing a point per game measurement, despite a weak correlation between junior scoring and NHL scoring for defenseman, is based off the research of Rhys Jessop. His research showcased that CHL defensemen who scored at a rate of 0.55 P/G or better in their draft season turn out to be much more likely at becoming a NHL defenseman. I took this research a step further by adjusting the P/G measurement based on the fact that power play scoring does not carry over to the NHL at the same rate as non-power play scoring. This was determined by looking at the career NHL season for CHL defensemen drafted between 1998 and 2011 that have played at least 250 NHL games. Overall, the power play P/G for these defenseman in their career NHL season is 51% of their power play P/G production in their draft year. Meanwhile, non-power play scoring, which includes even strength and shorthanded production, is 108% of their production in their draft year. Ultimately, the adjusted P/G measurement values a power play point at 47% compared with an even strength or shorthanded point. In my previous model, I utilized a cutoff of 0.44 P/G in a similar method to that of Rhys Jessop and only had two groupings. However, further research into the matter shows that method was flawed and the amount of groupings has been expanded to seven.

Similar to my method of grouping by draft position, I once again grouped the defensemen by finding where there are similar rates of return based on ATOI. For example, first year draft eligible CHL defensemen with an adjusted P/G between 0.40 and 0.53 have a similar rate of return. If you take a look at the scatter plot of the data it has a fairly flat trend line and a R² of 0.0003, meaning that for this sample, adjusted P/G explains only 0.03% of that player’s ATOI in the NHL. In terms of ATOI, there is basically no difference for defensemen that have an adjusted P/G between 0.40 and 0.53.

Ultimately, I divided the adjusted P/G into 7 groupings. Here are the results for games played for these groups:

In terms of games played, these charts demonstrate that the better the adjusted P/G production for CHL defenseman in their draft season the more likely they are to play more games at the NHL level.

Here are the results for ATOI for the same groupings:

Similar to games played, the better the adjusted P/G production for CHL defenseman in their draft year the better the chances of them playing more minutes at the NHL level.

COMBINING SCOUTING AND STATS

I have always been of the belief that the best way to predict the future success of a draft eligible player is to find where scouts and stats collide. For example, if we compare the average career ATOI of each of the 7 groupings, for the adjusted P/G measurement, against the groupings based on draft position we find that:

Adjusted P/G of 0.75+ is closest to a 3-6 overall pick.

Adjusted P/G of 0.66-0.74 is closest to a 7-9 overall pick.

Adjusted P/G of 0.54-0.65 is closest to a 20-29 overall pick.

Adjusted P/G of 0.40-0.53 is closest to a 30-56 overall pick.

Adjusted P/G of 0.19-0.39 is closest to a 70-108 overall pick.

Adjusted P/G of 0.11-0.18 is between an 109-162 pick and an 163-217 pick.

Adjusted P/G of 0.00-0.10 is worse than an 163-217 pick.

These results suggest that taking a high ranked defenseman in the top 9 with an adjusted P/G of 0.66 or greater should give you the best results and that is exactly what occurred.

CHL defenseman drafted in the top 9 that have an adjusted P/G of 0.66+ since 1998

Alex Pietrangelo 0.78 (24:40)

Jay Bouwmeester 0.76 (24:15)

Seth Jones 0.75 (22:31)

Zach Bogosian 0.73 (21:52)

Dougie Hamilton 0.73 (19:57)

Matthew Dumba 0.73 (19:25)

Ivan Provorov 0.72 (23:45)

Morgan Rielly 0.71 (21:24)

Brad Stuart 0.69 (21:24)

Aaron Ekblad 0.66 (22:26)

Average 0.73 (22:10)

CHL defenseman drafted in the top 9 that have an adjusted P/G of 0.54-0.65 since 1998

Mikhail Sergachev 0.65 (16:30)

Cam Barker 0.65 (17:07)

Derrick Pouliot 0.61 (17:09)

Drew Doughty 0.59 (26:17)

Olli Juolevi 0.59 (N/A)

Darnell Nurse 0.57 (21:21)

Ryan Murray 0.56 (20:23)

Erik Gudbranson 0.55 (18:05)

Rostislav Klesla 0.54 (20:02)

Average 0.59 (19:37)

CHL defenseman drafted in the top 9 that have an adjusted P/G of 0.40-0.53 since 1998

Haydn Fleury 0.53 (15:49)

Karl Alzner 0.52 (20:09)

Thomas Hickey 0.47 (18:07)

Griffin Reinhart 0.46 (17:14)

Luke Schenn 0.42 (18:16)

Dion Phaneuf 0.40 (23:28)

Average 0.47 (18:51)

CHL defenseman drafted in the top 9 that have an adjusted P/G of 0.19-0.39 since 1998

Jared Cowen 0.33 (18:53)

Bryan Allen 0.33 (18:38)

Braydon Coburn 0.27 (19:38)

Average 0.31 (19:03)

While there is value in both the eye test and in the numbers, neither should be looked at in a bubble. If you want to get the best results in drafting a defenseman you need to utilize both.