NHL Draft Value

[Buttonwood]

Something I'd like to look at someday (but honestly, it's not goaltender-related and therefore it always falls to the bottom of my list) is a comparison of the actual value of NHL draft picks (measured by some metric) versus the perceived value of NHL draft picks (measured by the value that NHL GMs impute when they trade those picks on or before draft day).

I'm throwing it up for grabs, and would love to be involved if things get moving.

I've actually toyed around with this. I 'predicted' career games of draft picks from 1996-2006 then divided that by Michael Schuckers draft value chart (derived for expected career games, http://myslu.stlawu.edu/~msch/sports/Schuckers_NHL_Draft.pdf) then multiplied it by a 'production factor' (ppg divided by average ppg). Summed them up by team year and volia!

Obviously, the weakest part of this assessment is the predicted games, but analyzing the draft and its impact pre-salary cap isn't as fun. Once I get around to it and tighten it up a bit, I'll probably pop it into a regression to explain team success (lagged 2-3 years of course).

Click to read more...

 

[Bear of Bad News]

I like this idea - and I wasn't actually aware that people had done research to impute the NHL teams' perceived value of draft picks.

For actual value, I was thinking of doing something using Hockey Reference's "point shares" as a proxy for value, discounting each year by some amount (8%, perhaps?) to reflect that a team loses the ability to control a player (free agency, et cetera) over time.

For instance, Trevor Linden - drafted second overall in 1988 - had 4.5 point shares his first season (so he gets full credit for that). His next year, he gets 92% credit for 3.2 point shares (or 2.9 point shares). Continuing the exercise for his entire career, I equate Linden's "draft value" at 43.2 DPS (discounted point shares).

Then you have to do this for all second overall picks (and then for all overall picks), and that's where I run out of time at the moment.

 

[Bear of Bad News]

Rounding out the top picks of the 1988 draft, I get.

#1 Mike Modano - 67.6 DPS
#2 Trevor Linden - 43.2 DPS
#3 Curtis Leschyshyn - 24.1 DPS
#4 Darrin Shannon - 9.9 DPS
#5 Daniel Dore - -0.1 DPS
#6 Scott Pearson - 2.6 DPS
#7 Martin Gelinas - 27.6 DPS
#8 Jeremy Roenick - 68.5 DPS
#9 Rod Brind'amour - 51.2 DPS
#10 Teemu Selanne - 64.6 DPS

Selanne gets penalized for two reasons - one, he started his great career a few years after the others, and two, he's probably still got value left in the reservoir. The estimation of the remaining reservoir is a challenge, but not an undefeatable one.

Shannon, Dore and Pearson bring up some interesting questions - should a player be penalized for a season of below 0 DPS? I'm leaning towards "no", or (backsliding somewhat) flooring the career DPS at 0.

 

[Buttonwood]

Early next week I'll try to determine how my (admittedly unsophisticated) formula would treat those picks. I would do it now, but the production factor would certainly be different for different eras and everything is on my work computer.

I like the idea of looking at 'draft value' as macro-level as possible. So I think some combination of games played, ice-time, and team points would be best. Franchise derive value out of wins, not necessarily scoring, and players that are part of those wins. Scoring represents value, but really only seems applicable to forwards. This would treat all 'win contributions' the same, defensive or offensive.

Point shares are certainly a step in the right direction though, I'll have to do some more research into it and then more importantly see if we can get the necessary data. The main reason I 'settled' on a formula based on games played, points, and draft position was the availability of data. (thanks, hockeyDB!) If we were to move towards a points share model, I would be tempted to re-do the denominator, draft pick value derived from career games. No small undertaking.

I think I'll re-do the chart with less emphasis on production (maybe make it standard deviation from average points per game). If there's is an available dataset of point shares or anything please let me know! The bigger the better.

 

[TasteofFlames]

Rounding out the top picks of the 1988 draft, I get.

#1 Mike Modano - 67.6 DPS
#2 Trevor Linden - 43.2 DPS
#3 Curtis Leschyshyn - 24.1 DPS
#4 Darrin Shannon - 9.9 DPS
#5 Daniel Dore - -0.1 DPS
#6 Scott Pearson - 2.6 DPS
#7 Martin Gelinas - 27.6 DPS
#8 Jeremy Roenick - 68.5 DPS
#9 Rod Brind'amour - 51.2 DPS
#10 Teemu Selanne - 64.6 DPS

Selanne gets penalized for two reasons - one, he started his great career a few years after the others, and two, he's probably still got value left in the reservoir. The estimation of the remaining reservoir is a challenge, but not an undefeatable one.

Shannon, Dore and Pearson bring up some interesting questions - should a player be penalized for a season of below 0 DPS? I'm leaning towards "no", or (backsliding somewhat) flooring the career DPS at 0.

Any particular reason Roenick ends up with a higher DPS than Modano or Selanne (you did put a couple of caveats on Selanne, so that's less surprising)? That just a case of no metric being perfect, or am I missing something in the calculation of these values?

 

[Bear of Bad News]

Roenick has 126 raw point shares (compared with Modano's 145 and Selanne's 168).

He does better by this metric because he had several 10/11-point seasons early in his career, while Modano's bulk came around the turn of the century, and Selanne didn't start accumulating his until 1992-93. The 8% discount factor really affects these results in particular.

Waiting for an afternoon event, I've plugged in 1988's draft and started on 1987. We'll see what I can come up with, given that I really should be finishing the database conversion on my website.

 

[Fourier]

This was interesting!

I must say that one of the main problems in determining how valuable a draft pick will be is quantifying what previous picks have actually contributed. The number of games played is a solid choice for a first kick at the problem, but of course as you point out it begs the question of how to further refine the quality of those games. In this respect I would say that without much to back up the claim I would say that the analysis does seem to significantly undervalue picks in the early part of the first round, but it is probably a pretty solid indicator of relative value for picks from say the middle of the second round onward.

I wonder how easy it would be in a community like this to actually come up with a scheme to refine the metrics by assigning some sort of "quality" weight to each players career. It may also be possible to look at recent salary data to help get a feel for those attributes of a prticular player that a team really values. After all we have to hope that most teams are paying for what the value most.

 

[Killion]

I wonder how easy it would be in a community like this to actually come up with a scheme to refine the metrics by assigning some sort of "quality" weight to each players career.

... through Binomial Distribution methodologies, practicable my dear Fourier. However, if your sampling is carried out without replacement, the draws wont be independent and you'll wind up with a Hypergeometric Distribution Sequence thus defeating the purpose of the exercise.

 

[Fourier]

... through Binomial Distribution methodologies, practicable my dear Fourier. However, if your sampling is carried out without replacement, the draws wont be independent and you'll wind up with a Hypergeometric Distribution Sequence thus defeating the purpose of the exercise.

Who let you in here! You may have the honour of being the forums first banned poster!

 

[Killion]

Who let you in here! You may have the honour of being the forums first banned poster!

... Most excellent. Held in Contempt of Court here at the Bournelli Trial. Disbarred. Disgraced. Just like Momma said Id wind up.

 

[Buttonwood]

Rounding out the top picks of the 1988 draft, I get.

#1 Mike Modano - 67.6 DPS
#2 Trevor Linden - 43.2 DPS
#3 Curtis Leschyshyn - 24.1 DPS
#4 Darrin Shannon - 9.9 DPS
#5 Daniel Dore - -0.1 DPS
#6 Scott Pearson - 2.6 DPS
#7 Martin Gelinas - 27.6 DPS
#8 Jeremy Roenick - 68.5 DPS
#9 Rod Brind'amour - 51.2 DPS
#10 Teemu Selanne - 64.6 DPS

Selanne gets penalized for two reasons - one, he started his great career a few years after the others, and two, he's probably still got value left in the reservoir. The estimation of the remaining reservoir is a challenge, but not an undefeatable one.

Shannon, Dore and Pearson bring up some interesting questions - should a player be penalized for a season of below 0 DPS? I'm leaning towards "no", or (backsliding somewhat) flooring the career DPS at 0.

Just for comparisons sake here is the same list using the simple formula (These players were measured against ppg in their own draft year, figured it was a more offensive time).

Adjusted Expected Games - Player

2,293 Mike Modano
1,447 Trevor Linden
721 Curtis Leschyshyn
417 Darrin Shannon
8 Daniel Dore
164 Scott Pearson
1,102 Martin Gelinas
2,030 Jeremy Roenick
1,976 Rod Brind'Amour
2,347 Teemu Selanne

Not horrible, not great either. I'm going to change the 'production factor' slightly to not give so much emphasis to offense. I'll re-post the team year table and maybe a draft just for arguments sake so people can pick it apart.

And yes, I'm a Flames fan so obviously Martin Gelinas was almost as good as Linden.