a look at draft by numbers: using data and analytics to improve...
TRANSCRIPT
![Page 1: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/1.jpg)
A Look at Draft By Numbers: Using Data andAnalytics to Improve NHL Player Selection
Michael Schuckers
St. Lawrence Univ., Canton, NYStatistical Sports Consulting, LLC
@SchuckersM
@EmpiricalSports
May 2, 2020
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 1 / 35
![Page 2: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/2.jpg)
Hockey Analytics Night in Where?
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 2 / 35
![Page 3: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/3.jpg)
Motivation & Goals
NHL Player Selection
If you look at statistics and point to a column and say, ’We’redrafting this guy’ — have fun. I hope you’re in my division.- Brian Burke, MIT Sloan Sports Analytics Conference, 2013
Source: Article by Dave Feschuk, The Star, March 1, 2013
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 3 / 35
![Page 4: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/4.jpg)
Motivation & Goals
Primary Goals
Paper Take historical data publicly available to teams at the time ofthe draft and build a model to compare with how teams did. Can wedraft better than team’s with just analytics?
Talk Walk through some of the methods used in the paper,particularly Generalized Additive Models (GAM’s) and interactions.
Version of the Paper
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 4 / 35
![Page 5: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/5.jpg)
Motivation & Goals
Primary Goals
Paper Take historical data publicly available to teams at the time ofthe draft and build a model to compare with how teams did. Can wedraft better than team’s with just analytics?
Talk Walk through some of the methods used in the paper,particularly Generalized Additive Models (GAM’s) and interactions.
Version of the Paper
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 4 / 35
![Page 6: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/6.jpg)
Motivation & Goals
Primary Goals
Paper Take historical data publicly available to teams at the time ofthe draft and build a model to compare with how teams did. Can wedraft better than team’s with just analytics?
Talk Walk through some of the methods used in the paper,particularly Generalized Additive Models (GAM’s) and interactions.
Version of the Paper
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 4 / 35
![Page 7: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/7.jpg)
Motivation & Goals
Primary Goals
Paper Take historical data publicly available to teams at the time ofthe draft and build a model to compare with how teams did. Can wedraft better than team’s with just analytics?
Talk Walk through some of the methods used in the paper,particularly Generalized Additive Models (GAM’s) and interactions.
Version of the Paper
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 4 / 35
![Page 8: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/8.jpg)
Motivation & Goals
NHL Entry Draft
Annually at the end of June over two days
Method by which newly eligible players are allocated to NHL teams
Eligible players: 18 years old on or before September 15 and not olderthan 20 years old before December 31 (Worldwide)
Each team (of 31) starts with a pick in each of 7 rounds; picks aretradeable
Team with the worst record has best chance at first pick in first round(lottery among non-playoff teams)
After 1st round, picks are made in order of reg. season finish
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 5 / 35
![Page 9: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/9.jpg)
Motivation
Motivation the 2011 NHL Draft
The 2011 NHL Draft, first six picks
1 EDM: Ryan Nugent-Hopkins (F)
2 COL: Gabriel Landeskog (F)
3 FLA: Jonathan Huberdeau (F)
4 NJD: Adam Larsson (D)
5 NYI: Ryan Strome (F)
6 OTT: Mika Zibanejad (F)
7 WPG: ????
Who should WPG select? Consensus Available (in order):Sean Couturier (F), Dougie Hamilton (D), Ryan Murphy (D),Sven Baertschi (F), Nathan Beaulieu (D), Duncan Siemens (D), MarkScheifele (F)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 6 / 35
![Page 10: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/10.jpg)
Motivation
Motivation the 2011 NHL Draft
The 2011 NHL Draft, first six picks
1 EDM: Ryan Nugent-Hopkins (F)
2 COL: Gabriel Landeskog (F)
3 FLA: Jonathan Huberdeau (F)
4 NJD: Adam Larsson (D)
5 NYI: Ryan Strome (F)
6 OTT: Mika Zibanejad (F)
7 WPG: ????
Who should WPG select? Consensus Available (in order):Sean Couturier (F), Dougie Hamilton (D), Ryan Murphy (D),Sven Baertschi (F), Nathan Beaulieu (D), Duncan Siemens (D), MarkScheifele (F)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 6 / 35
![Page 11: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/11.jpg)
Motivation
Motivation the 2011 NHL Draft
The 2011 NHL Draft, first six picks
1 EDM: Ryan Nugent-Hopkins (F)
2 COL: Gabriel Landeskog (F)
3 FLA: Jonathan Huberdeau (F)
4 NJD: Adam Larsson (D)
5 NYI: Ryan Strome (F)
6 OTT: Mika Zibanejad (F)
7 WPG: ????
Who should WPG select? Consensus Available (in order):Sean Couturier (F), Dougie Hamilton (D), Ryan Murphy (D),Sven Baertschi (F), Nathan Beaulieu (D), Duncan Siemens (D), MarkScheifele (F)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 6 / 35
![Page 12: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/12.jpg)
Motivation
Motivation the 2011 NHL Draft
The 2011 NHL Draft, first six picks
1 EDM: Ryan Nugent-Hopkins (F)
2 COL: Gabriel Landeskog (F)
3 FLA: Jonathan Huberdeau (F)
4 NJD: Adam Larsson (D)
5 NYI: Ryan Strome (F)
6 OTT: Mika Zibanejad (F)
7 WPG: ????
Who should WPG select? Consensus Available (in order):Sean Couturier (F), Dougie Hamilton (D), Ryan Murphy (D),Sven Baertschi (F), Nathan Beaulieu (D), Duncan Siemens (D), MarkScheifele (F)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 6 / 35
![Page 13: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/13.jpg)
Motivation
Motivation the 2011 NHL Draft
The 2011 NHL Draft, first six picks
1 EDM: Ryan Nugent-Hopkins (F)
2 COL: Gabriel Landeskog (F)
3 FLA: Jonathan Huberdeau (F)
4 NJD: Adam Larsson (D)
5 NYI: Ryan Strome (F)
6 OTT: Mika Zibanejad (F)
7 WPG: ????
Who should WPG select? Consensus Available (in order):Sean Couturier (F), Dougie Hamilton (D), Ryan Murphy (D),Sven Baertschi (F), Nathan Beaulieu (D), Duncan Siemens (D), MarkScheifele (F)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 6 / 35
![Page 14: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/14.jpg)
Motivation
Motivation the 2011 NHL Draft
The 2011 NHL Draft, first six picks
1 EDM: Ryan Nugent-Hopkins (F)
2 COL: Gabriel Landeskog (F)
3 FLA: Jonathan Huberdeau (F)
4 NJD: Adam Larsson (D)
5 NYI: Ryan Strome (F)
6 OTT: Mika Zibanejad (F)
7 WPG: ????
Who should WPG select? Consensus Available (in order):Sean Couturier (F), Dougie Hamilton (D), Ryan Murphy (D),Sven Baertschi (F), Nathan Beaulieu (D), Duncan Siemens (D), MarkScheifele (F)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 6 / 35
![Page 15: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/15.jpg)
Motivation
Motivation the 2011 NHL Draft
The 2011 NHL Draft, first six picks
1 EDM: Ryan Nugent-Hopkins (F)
2 COL: Gabriel Landeskog (F)
3 FLA: Jonathan Huberdeau (F)
4 NJD: Adam Larsson (D)
5 NYI: Ryan Strome (F)
6 OTT: Mika Zibanejad (F)
7 WPG: ????
Who should WPG select? Consensus Available (in order):Sean Couturier (F), Dougie Hamilton (D), Ryan Murphy (D),Sven Baertschi (F), Nathan Beaulieu (D), Duncan Siemens (D), MarkScheifele (F)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 6 / 35
![Page 16: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/16.jpg)
Motivation
Motivation the 2011 NHL Draft
The 2011 NHL Draft, first six picks
1 EDM: Ryan Nugent-Hopkins (F)
2 COL: Gabriel Landeskog (F)
3 FLA: Jonathan Huberdeau (F)
4 NJD: Adam Larsson (D)
5 NYI: Ryan Strome (F)
6 OTT: Mika Zibanejad (F)
7 WPG: ????
Who should WPG select? Consensus Available (in order):
Sean Couturier (F), Dougie Hamilton (D), Ryan Murphy (D),Sven Baertschi (F), Nathan Beaulieu (D), Duncan Siemens (D), MarkScheifele (F)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 6 / 35
![Page 17: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/17.jpg)
Motivation
Motivation the 2011 NHL Draft
The 2011 NHL Draft, first six picks
1 EDM: Ryan Nugent-Hopkins (F)
2 COL: Gabriel Landeskog (F)
3 FLA: Jonathan Huberdeau (F)
4 NJD: Adam Larsson (D)
5 NYI: Ryan Strome (F)
6 OTT: Mika Zibanejad (F)
7 WPG: ????
Who should WPG select? Consensus Available (in order):
Sean Couturier (F), Dougie Hamilton (D), Ryan Murphy (D),Sven Baertschi (F), Nathan Beaulieu (D), Duncan Siemens (D), MarkScheifele (F)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 6 / 35
![Page 18: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/18.jpg)
Motivation
Motivation the 2011 NHL Draft
The 2011 NHL Draft, first six picks
1 EDM: Ryan Nugent-Hopkins (F)
2 COL: Gabriel Landeskog (F)
3 FLA: Jonathan Huberdeau (F)
4 NJD: Adam Larsson (D)
5 NYI: Ryan Strome (F)
6 OTT: Mika Zibanejad (F)
7 WPG: ????
Who should WPG select? Consensus Available (in order):Sean Couturier (F), Dougie Hamilton (D), Ryan Murphy (D),Sven Baertschi (F), Nathan Beaulieu (D), Duncan Siemens (D), MarkScheifele (F)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 6 / 35
![Page 19: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/19.jpg)
Motivation
Motivation the 2011 NHL Draft
The 2011 NHL Draft, first six picks
1 EDM: Ryan Nugent-Hopkins (F)
2 COL: Gabriel Landeskog (F)
3 FLA: Jonathan Huberdeau (F)
4 NJD: Adam Larsson (D)
5 NYI: Ryan Strome (F)
6 OTT: Mika Zibanejad (F)
7 WPG: Mark Scheifele (F)
Who should WPG select? Consensus Available (in order):Sean Couturier (F), Dougie Hamilton (D), Ryan Murphy (D),Sven Baertschi (F), Nathan Beaulieu (D), Duncan Siemens (D), MarkScheifele (F)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 7 / 35
![Page 20: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/20.jpg)
Motivation
Motivation the 2011 NHL Draft
The 2011 NHL Draft, first six picks
1 EDM: Ryan Nugent-Hopkins (F)
2 COL: Gabriel Landeskog (F)
3 FLA: Jonathan Huberdeau (F)
4 NJD: Adam Larsson (D)
5 NYI: Ryan Strome (F)
6 OTT: Mika Zibanejad (F)
7 WPG: Mark Scheifele (F)
8 PHL: Sean Couturier (F)
9 BOS: Dougie Hamilton (D)
10 MIN: Jonas Brodin (D)
11 COL: Duncan Siemens (D)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 8 / 35
![Page 21: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/21.jpg)
Motivation
Motivation the 2011 NHL Draft
Consensus that Scheifele pick was a moderate surprise
Some analytics about Couturier >> Scheifele
Post draft: Couturier had 77, 46 GP (Games Played) 1st 2 seasons
Post draft: Scheifele had 7, 4 GP 1st 2 seasons
Today after 8+ seasons: Couturier has 647 GP, 402 Pts, 156 Goals
Today after 8+ seasons: Scheifele has 519 GP, 444 Pts, 180 Goals
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 9 / 35
![Page 22: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/22.jpg)
Motivation
Motivation the 2011 NHL Draft
Consensus that Scheifele pick was a moderate surprise
Some analytics about Couturier >> Scheifele
Post draft: Couturier had 77, 46 GP (Games Played) 1st 2 seasons
Post draft: Scheifele had 7, 4 GP 1st 2 seasons
Today after 8+ seasons: Couturier has 647 GP, 402 Pts, 156 Goals
Today after 8+ seasons: Scheifele has 519 GP, 444 Pts, 180 Goals
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 9 / 35
![Page 23: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/23.jpg)
Motivation
Motivation the 2011 NHL Draft
Consensus that Scheifele pick was a moderate surprise
Some analytics about Couturier >> Scheifele
Post draft: Couturier had 77, 46 GP (Games Played) 1st 2 seasons
Post draft: Scheifele had 7, 4 GP 1st 2 seasons
Today after 8+ seasons: Couturier has 647 GP, 402 Pts, 156 Goals
Today after 8+ seasons: Scheifele has 519 GP, 444 Pts, 180 Goals
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 9 / 35
![Page 24: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/24.jpg)
Motivation
Motivation the 2011 NHL Draft
Consensus that Scheifele pick was a moderate surprise
Some analytics about Couturier >> Scheifele
Post draft: Couturier had 77, 46 GP (Games Played) 1st 2 seasons
Post draft: Scheifele had 7, 4 GP 1st 2 seasons
Today after 8+ seasons: Couturier has 647 GP, 402 Pts, 156 Goals
Today after 8+ seasons: Scheifele has 519 GP, 444 Pts, 180 Goals
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 9 / 35
![Page 25: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/25.jpg)
Motivation
Motivation the 2011 NHL Draft
Consensus that Scheifele pick was a moderate surprise
Some analytics about Couturier >> Scheifele
Post draft: Couturier had 77, 46 GP (Games Played) 1st 2 seasons
Post draft: Scheifele had 7, 4 GP 1st 2 seasons
Today after 8+ seasons: Couturier has 647 GP, 402 Pts, 156 Goals
Today after 8+ seasons: Scheifele has 519 GP, 444 Pts, 180 Goals
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 9 / 35
![Page 26: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/26.jpg)
Motivation
Motivation the 2011 NHL Draft
Consensus that Scheifele pick was a moderate surprise
Some analytics about Couturier >> Scheifele
Post draft: Couturier had 77, 46 GP (Games Played) 1st 2 seasons
Post draft: Scheifele had 7, 4 GP 1st 2 seasons
Today after 8+ seasons: Couturier has 647 GP, 402 Pts, 156 Goals
Today after 8+ seasons: Scheifele has 519 GP, 444 Pts, 180 Goals
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 9 / 35
![Page 27: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/27.jpg)
Motivation
Motivation the 2011 NHL Draft
Illustrate some themes:
Evaluation of players is hard
and it takes time (players develop at different rates from 18yo)
Metrics for assessment are not simple
nor single agreed response metric
In short, noisy and sparse data with slow feedback and
(to teams) results incredibly important
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 10 / 35
![Page 28: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/28.jpg)
Motivation
Motivation the 2011 NHL Draft
Illustrate some themes:
Evaluation of players is hard
and it takes time (players develop at different rates from 18yo)
Metrics for assessment are not simple
nor single agreed response metric
In short, noisy and sparse data with slow feedback and
(to teams) results incredibly important
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 10 / 35
![Page 29: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/29.jpg)
Motivation
Motivation the 2011 NHL Draft
Illustrate some themes:
Evaluation of players is hard
and it takes time (players develop at different rates from 18yo)
Metrics for assessment are not simple
nor single agreed response metric
In short, noisy and sparse data with slow feedback and
(to teams) results incredibly important
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 10 / 35
![Page 30: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/30.jpg)
Motivation
Motivation the 2011 NHL Draft
Illustrate some themes:
Evaluation of players is hard
and it takes time (players develop at different rates from 18yo)
Metrics for assessment are not simple
nor single agreed response metric
In short, noisy and sparse data with slow feedback and
(to teams) results incredibly important
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 10 / 35
![Page 31: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/31.jpg)
Motivation
Motivation the 2011 NHL Draft
Illustrate some themes:
Evaluation of players is hard
and it takes time (players develop at different rates from 18yo)
Metrics for assessment are not simple
nor single agreed response metric
In short, noisy and sparse data with slow feedback and
(to teams) results incredibly important
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 10 / 35
![Page 32: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/32.jpg)
Motivation
Motivation the 2011 NHL Draft
Illustrate some themes:
Evaluation of players is hard
and it takes time (players develop at different rates from 18yo)
Metrics for assessment are not simple
nor single agreed response metric
In short, noisy and sparse data with slow feedback and
(to teams) results incredibly important
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 10 / 35
![Page 33: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/33.jpg)
Motivation
NHL Player Selection
Desjardins (2004-), NHL League Equivalencies (NHLe),
Estimate ρ` where YNHL,t = ρ`Y`,t−1
ratio estimator
for each league, `, e.g. SEL, OHL, QMJHL, NCAA
where Y is Points generally
t is year, t − 1 previous year
can be adjusted further by TOI
allows a measure of league quality
need amount of data
Extended by Rob Vollman, added by age ρ`,age
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 11 / 35
![Page 34: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/34.jpg)
Motivation
NHLe from 2014-15
source:
http://www.hockeyabstract.com/thoughts/updatedtranslationfactors
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 12 / 35
![Page 35: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/35.jpg)
Motivation
NHL Player Selection
Lawrence and Weissbock (2015), Prospect Cohort Success
Nearest neighbors approach
Inputs: League, Age, Points, Height
Generated comparable players (cohort)
Pct Success (> 200 GP) among cohort
Out of sample testing outperformed teams
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 13 / 35
![Page 36: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/36.jpg)
Data
Sources of Information
Response/Outcome metrics (eg TOI in NHL)
Demographics (eg Height and Weight)
Pre-Draft Performance (eg PPG, SVPct)
Scouting (via Central Scouting Service)
Evaluation of draft picks takes time esp. in Hockey
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 14 / 35
![Page 37: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/37.jpg)
Data
Sources of Information
Response/Outcome metrics (eg TOI in NHL)
Demographics (eg Height and Weight)
Pre-Draft Performance (eg PPG, SVPct)
Scouting (via Central Scouting Service)
Evaluation of draft picks takes time esp. in Hockey
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 14 / 35
![Page 38: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/38.jpg)
Data
Sources of Information
Response/Outcome metrics (eg TOI in NHL)
Demographics (eg Height and Weight)
Pre-Draft Performance (eg PPG, SVPct)
Scouting (via Central Scouting Service)
Evaluation of draft picks takes time esp. in Hockey
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 14 / 35
![Page 39: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/39.jpg)
Data
Sources of Information
Response/Outcome metrics (eg TOI in NHL)
Demographics (eg Height and Weight)
Pre-Draft Performance (eg PPG, SVPct)
Scouting (via Central Scouting Service)
Evaluation of draft picks takes time esp. in Hockey
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 14 / 35
![Page 40: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/40.jpg)
Data
Sources of Information
Response/Outcome metrics (eg TOI in NHL)
Demographics (eg Height and Weight)
Pre-Draft Performance (eg PPG, SVPct)
Scouting (via Central Scouting Service)
Evaluation of draft picks takes time esp. in Hockey
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 14 / 35
![Page 41: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/41.jpg)
Data
Sources of Information
Response/Outcome metrics (eg TOI in NHL)
Demographics (eg Height and Weight)
Pre-Draft Performance (eg PPG, SVPct)
Scouting (via Central Scouting Service)
Evaluation of draft picks takes time esp. in Hockey
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 14 / 35
![Page 42: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/42.jpg)
Data
Response metrics
Goals:
Available for all players in corpus
Relevant to performance
Comparable across positions
Length of time? Career? First 5 Years?
Choices:
Time on Ice (TOI)
Games Played (GP)
Cumulative First Seven Years per CBA (Schuckers and Argeris, 2015)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 15 / 35
![Page 43: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/43.jpg)
Data
Response metrics
Goals:
Available for all players in corpus
Relevant to performance
Comparable across positions
Length of time? Career? First 5 Years?
Choices:
Time on Ice (TOI)
Games Played (GP)
Cumulative First Seven Years per CBA (Schuckers and Argeris, 2015)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 15 / 35
![Page 44: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/44.jpg)
Data
Response metrics
Goals:
Available for all players in corpus
Relevant to performance
Comparable across positions
Length of time? Career? First 5 Years?
Choices:
Time on Ice (TOI)
Games Played (GP)
Cumulative First Seven Years per CBA (Schuckers and Argeris, 2015)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 15 / 35
![Page 45: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/45.jpg)
Data
Response metrics
Goals:
Available for all players in corpus
Relevant to performance
Comparable across positions
Length of time? Career? First 5 Years?
Choices:
Time on Ice (TOI)
Games Played (GP)
Cumulative First Seven Years per CBA (Schuckers and Argeris, 2015)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 15 / 35
![Page 46: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/46.jpg)
Data
Response metrics
Goals:
Available for all players in corpus
Relevant to performance
Comparable across positions
Length of time? Career? First 5 Years?
Choices:
Time on Ice (TOI)
Games Played (GP)
Cumulative First Seven Years per CBA (Schuckers and Argeris, 2015)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 15 / 35
![Page 47: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/47.jpg)
Data
Response metrics
Goals:
Available for all players in corpus
Relevant to performance
Comparable across positions
Length of time? Career? First 5 Years?
Choices:
Time on Ice (TOI)
Games Played (GP)
Cumulative First Seven Years per CBA (Schuckers and Argeris, 2015)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 15 / 35
![Page 48: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/48.jpg)
Data
Response metrics
Goals:
Available for all players in corpus
Relevant to performance
Comparable across positions
Length of time? Career? First 5 Years?
Choices:
Time on Ice (TOI)
Games Played (GP)
Cumulative First Seven Years per CBA (Schuckers and Argeris, 2015)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 15 / 35
![Page 49: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/49.jpg)
Data
Response metrics
Goals:
Available for all players in corpus
Relevant to performance
Comparable across positions
Length of time? Career? First 5 Years?
Choices:
Time on Ice (TOI)
Games Played (GP)
Cumulative First Seven Years per CBA (Schuckers and Argeris, 2015)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 15 / 35
![Page 50: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/50.jpg)
Data
Demographics
Choices:
Height
Weight
Position (C, F, D, G)
and functions of these.As much as possible taken from draft eligible window.
In this NHL.com is pretty good because they don’t update theirsite.
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 16 / 35
![Page 51: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/51.jpg)
Data
Pre-Draft Performance
Needs:
Data has to be available for players in corpus
Available for nearly every league in draft -1 years
Need to know league drafted from
Choices:
PPG
GAA
GP
Leagues(Liiga, NCAA, OHL, Other, QMJHL, Russia, Russia2, USHL,WHL)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 17 / 35
![Page 52: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/52.jpg)
Data
Scouting Data
Via Central Scouting Player Rankings
NHL Internal Rankings from Central Scouting Service
Four Groups: (Skaters, Goalies) × (North Amer., Europe)
Convert to single list via Iain Fyffe’s Cescin (2011)
Central Scouting Integrator (Cescin)
NA Skaters = 1.35, NA Goalies = 13.25EU Skaters = 6.27, EU Goalies = 38.18
To get Cescin multiply values from above by CSS Rank
J Quinn 2020 #9 Ranked NA Skater would be CESCIN= 9 × 1.35 = 12.15
Link to Fyffe Article, Link to 2020 CSS Rankings
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 18 / 35
![Page 53: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/53.jpg)
Data
Scouting Data
Via Central Scouting Player Rankings
NHL Internal Rankings from Central Scouting Service
Four Groups: (Skaters, Goalies) × (North Amer., Europe)
Convert to single list via Iain Fyffe’s Cescin (2011)
Central Scouting Integrator (Cescin)
NA Skaters = 1.35, NA Goalies = 13.25EU Skaters = 6.27, EU Goalies = 38.18
To get Cescin multiply values from above by CSS Rank
J Quinn 2020 #9 Ranked NA Skater would be CESCIN= 9 × 1.35 = 12.15
Link to Fyffe Article, Link to 2020 CSS Rankings
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 18 / 35
![Page 54: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/54.jpg)
Data
Scouting Data
Via Central Scouting Player Rankings
NHL Internal Rankings from Central Scouting Service
Four Groups: (Skaters, Goalies) × (North Amer., Europe)
Convert to single list via Iain Fyffe’s Cescin (2011)
Central Scouting Integrator (Cescin)
NA Skaters = 1.35, NA Goalies = 13.25EU Skaters = 6.27, EU Goalies = 38.18
To get Cescin multiply values from above by CSS Rank
J Quinn 2020 #9 Ranked NA Skater would be CESCIN= 9 × 1.35 = 12.15
Link to Fyffe Article, Link to 2020 CSS Rankings
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 18 / 35
![Page 55: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/55.jpg)
Data
Scouting Data
Via Central Scouting Player Rankings
NHL Internal Rankings from Central Scouting Service
Four Groups: (Skaters, Goalies) × (North Amer., Europe)
Convert to single list via Iain Fyffe’s Cescin (2011)
Central Scouting Integrator (Cescin)
NA Skaters = 1.35, NA Goalies = 13.25EU Skaters = 6.27, EU Goalies = 38.18
To get Cescin multiply values from above by CSS Rank
J Quinn 2020 #9 Ranked NA Skater would be CESCIN= 9 × 1.35 = 12.15
Link to Fyffe Article, Link to 2020 CSS Rankings
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 18 / 35
![Page 56: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/56.jpg)
Data
Scouting Data
Via Central Scouting Player Rankings
NHL Internal Rankings from Central Scouting Service
Four Groups: (Skaters, Goalies) × (North Amer., Europe)
Convert to single list via Iain Fyffe’s Cescin (2011)
Central Scouting Integrator (Cescin)
NA Skaters = 1.35, NA Goalies = 13.25
EU Skaters = 6.27, EU Goalies = 38.18
To get Cescin multiply values from above by CSS Rank
J Quinn 2020 #9 Ranked NA Skater would be CESCIN= 9 × 1.35 = 12.15
Link to Fyffe Article, Link to 2020 CSS Rankings
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 18 / 35
![Page 57: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/57.jpg)
Data
Scouting Data
Via Central Scouting Player Rankings
NHL Internal Rankings from Central Scouting Service
Four Groups: (Skaters, Goalies) × (North Amer., Europe)
Convert to single list via Iain Fyffe’s Cescin (2011)
Central Scouting Integrator (Cescin)
NA Skaters = 1.35, NA Goalies = 13.25EU Skaters = 6.27, EU Goalies = 38.18
To get Cescin multiply values from above by CSS Rank
J Quinn 2020 #9 Ranked NA Skater would be CESCIN= 9 × 1.35 = 12.15
Link to Fyffe Article, Link to 2020 CSS Rankings
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 18 / 35
![Page 58: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/58.jpg)
Data
Scouting Data
Via Central Scouting Player Rankings
NHL Internal Rankings from Central Scouting Service
Four Groups: (Skaters, Goalies) × (North Amer., Europe)
Convert to single list via Iain Fyffe’s Cescin (2011)
Central Scouting Integrator (Cescin)
NA Skaters = 1.35, NA Goalies = 13.25EU Skaters = 6.27, EU Goalies = 38.18
To get Cescin multiply values from above by CSS Rank
J Quinn 2020 #9 Ranked NA Skater would be CESCIN= 9 × 1.35 = 12.15
Link to Fyffe Article, Link to 2020 CSS Rankings
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 18 / 35
![Page 59: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/59.jpg)
Data
Scouting Data
Via Central Scouting Player Rankings
NHL Internal Rankings from Central Scouting Service
Four Groups: (Skaters, Goalies) × (North Amer., Europe)
Convert to single list via Iain Fyffe’s Cescin (2011)
Central Scouting Integrator (Cescin)
NA Skaters = 1.35, NA Goalies = 13.25EU Skaters = 6.27, EU Goalies = 38.18
To get Cescin multiply values from above by CSS Rank
J Quinn 2020 #9 Ranked NA Skater would be CESCIN= 9 × 1.35 = 12.15
Link to Fyffe Article, Link to 2020 CSS Rankings
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 18 / 35
![Page 60: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/60.jpg)
Data
Scouting Data
Via Central Scouting Player Rankings
NHL Internal Rankings from Central Scouting Service
Four Groups: (Skaters, Goalies) × (North Amer., Europe)
Convert to single list via Iain Fyffe’s Cescin (2011)
Central Scouting Integrator (Cescin)
NA Skaters = 1.35, NA Goalies = 13.25EU Skaters = 6.27, EU Goalies = 38.18
To get Cescin multiply values from above by CSS Rank
J Quinn 2020 #9 Ranked NA Skater would be CESCIN= 9 × 1.35 = 12.15
Link to Fyffe Article, Link to 2020 CSS Rankings
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 18 / 35
![Page 61: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/61.jpg)
Data
Scouting Data
Via Central Scouting Player Rankings
NHL Internal Rankings from Central Scouting Service
Four Groups: (Skaters, Goalies) × (North Amer., Europe)
Convert to single list via Iain Fyffe’s Cescin (2011)
Central Scouting Integrator (Cescin)
NA Skaters = 1.35, NA Goalies = 13.25EU Skaters = 6.27, EU Goalies = 38.18
To get Cescin multiply values from above by CSS Rank
J Quinn 2020 #9 Ranked NA Skater would be CESCIN= 9 × 1.35 = 12.15
Link to Fyffe Article, Link to 2020 CSS Rankings
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 18 / 35
![Page 62: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/62.jpg)
Data
Join the Data
Large Data Wrangling Effort
Combine Data by Player & Draft Year
Some players ranked by CSS but not drafted
Data in two cohorts
1998-00 (Training), 2001-02 (Test), 2007-08 (Validate)2004-06 (Training), 2007-08 (Test)
Spellings, accents, multiple players same name
I was inefficient at this task at best.
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 19 / 35
![Page 63: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/63.jpg)
Data
Robin Olssons (Born in ’89 or ’90)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 20 / 35
![Page 64: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/64.jpg)
Model
Model
Warning: Notation ahead
g−1(Yi ) = β0 + β1f1(X1i ) + β2f2(X2i ) + . . .+ βk fk(Xki )
Predict Response Yi (either 1st 7 GP or 1st 7 TOI) for player i
g() is link function, we will use g() = log() via Poisson family
Generalized Additive Model
fj()’s variety of functional forms and fits (Splines, Loess, quadratic)
More flexible relative to regular multiple linear regression
Still linear on the Right Hand Side
And interactions
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 21 / 35
![Page 65: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/65.jpg)
Model
Model
Warning: Notation ahead
g−1(Yi ) = β0 + β1f1(X1i ) + β2f2(X2i ) + . . .+ βk fk(Xki )
Predict Response Yi (either 1st 7 GP or 1st 7 TOI) for player i
g() is link function, we will use g() = log() via Poisson family
Generalized Additive Model
fj()’s variety of functional forms and fits (Splines, Loess, quadratic)
More flexible relative to regular multiple linear regression
Still linear on the Right Hand Side
And interactions
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 21 / 35
![Page 66: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/66.jpg)
Model
Model
Warning: Notation ahead
g−1(Yi ) = β0 + β1f1(X1i ) + β2f2(X2i ) + . . .+ βk fk(Xki )
Predict Response Yi (either 1st 7 GP or 1st 7 TOI) for player i
g() is link function, we will use g() = log() via Poisson family
Generalized Additive Model
fj()’s variety of functional forms and fits (Splines, Loess, quadratic)
More flexible relative to regular multiple linear regression
Still linear on the Right Hand Side
And interactions
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 21 / 35
![Page 67: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/67.jpg)
Model
Model
Warning: Notation ahead
g−1(Yi ) = β0 + β1f1(X1i ) + β2f2(X2i ) + . . .+ βk fk(Xki )
Predict Response Yi (either 1st 7 GP or 1st 7 TOI) for player i
g() is link function, we will use g() = log() via Poisson family
Generalized Additive Model
fj()’s variety of functional forms and fits (Splines, Loess, quadratic)
More flexible relative to regular multiple linear regression
Still linear on the Right Hand Side
And interactions
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 21 / 35
![Page 68: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/68.jpg)
Model
Model
Warning: Notation ahead
g−1(Yi ) = β0 + β1f1(X1i ) + β2f2(X2i ) + . . .+ βk fk(Xki )
Predict Response Yi (either 1st 7 GP or 1st 7 TOI) for player i
g() is link function, we will use g() = log() via Poisson family
Generalized Additive Model
fj()’s variety of functional forms and fits (Splines, Loess, quadratic)
More flexible relative to regular multiple linear regression
Still linear on the Right Hand Side
And interactions
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 21 / 35
![Page 69: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/69.jpg)
Model
Model
Warning: Notation ahead
g−1(Yi ) = β0 + β1f1(X1i ) + β2f2(X2i ) + . . .+ βk fk(Xki )
Predict Response Yi (either 1st 7 GP or 1st 7 TOI) for player i
g() is link function, we will use g() = log() via Poisson family
Generalized Additive Model
fj()’s variety of functional forms and fits (Splines, Loess, quadratic)
More flexible relative to regular multiple linear regression
Still linear on the Right Hand Side
And interactions
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 21 / 35
![Page 70: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/70.jpg)
Model
Model
Warning: Notation ahead
g−1(Yi ) = β0 + β1f1(X1i ) + β2f2(X2i ) + . . .+ βk fk(Xki )
Predict Response Yi (either 1st 7 GP or 1st 7 TOI) for player i
g() is link function, we will use g() = log() via Poisson family
Generalized Additive Model
fj()’s variety of functional forms and fits (Splines, Loess, quadratic)
More flexible relative to regular multiple linear regression
Still linear on the Right Hand Side
And interactions
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 21 / 35
![Page 71: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/71.jpg)
Model
Model
Warning: Notation ahead
g−1(Yi ) = β0 + β1f1(X1i ) + β2f2(X2i ) + . . .+ βk fk(Xki )
Predict Response Yi (either 1st 7 GP or 1st 7 TOI) for player i
g() is link function, we will use g() = log() via Poisson family
Generalized Additive Model
fj()’s variety of functional forms and fits (Splines, Loess, quadratic)
More flexible relative to regular multiple linear regression
Still linear on the Right Hand Side
And interactions
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 21 / 35
![Page 72: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/72.jpg)
Model
Model
Warning: Notation ahead
g−1(Yi ) = β0 + β1f1(X1i ) + β2f2(X2i ) + . . .+ βk fk(Xki )
Predict Response Yi (either 1st 7 GP or 1st 7 TOI) for player i
g() is link function, we will use g() = log() via Poisson family
Generalized Additive Model
fj()’s variety of functional forms and fits (Splines, Loess, quadratic)
More flexible relative to regular multiple linear regression
Still linear on the Right Hand Side
And interactions
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 21 / 35
![Page 73: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/73.jpg)
Model
GAM’s
Relationship between GP and Cescin
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 22 / 35
![Page 74: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/74.jpg)
Model
GAM’s
Multiple Regression
Yi = β0 + β1X1i + β2X2i + . . .+ βkXki
GLM
g−1(Yi ) = β0 + β1X1i + β2X2i + . . .+ βkXki
GAM
g−1(Yi ) = β0 + β1f1(X1i ) + β2f2(X2i ) + . . .+ βk fk(Xki )
NB: There are other differences in estimation for non-Gaussian families
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 23 / 35
![Page 75: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/75.jpg)
Model
Additivity
Typical Linear Multiple Regression
Yi = β0 + β1X1i + β2X2i + . . .+ βkXki
Additivity means each effect adds to the others.Transformation (via g() or g−1())such as log or logistic means that theimpacts are different. In particular, log transformation means amultiplicative effect.
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 24 / 35
![Page 76: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/76.jpg)
Model
Interactions and Indicator Variables
Interactions and Indicators allow for nuanced models.Example: Indicator Variables
xi =
{1 if player i is a Goalie,0 if player i is not a Goalie
(1)
Interactions are variables created by multiplying two variables together.Eg. X7 = X3X4 but allows for great model flexibility
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 25 / 35
![Page 77: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/77.jpg)
Model
Interactions: Example
IG , IF , ID Indicator (Positions), IQMJHL Indicator (QMJHL)
Full (Grossly Hypothetical) Model:
TOI ∼ 400 + 60IG + 0.01Height + 0.2PPG × IF + 0.25Height × IG
− 0.005 × f (Cescin) − 0.03ID × IQMJHL
For G:TOI ∼ 400 + 60 + (0.01 + 0.25)Height − 0.005f (Cescin)
For F:TOI ∼ 400 + 0.01Height + 0.2PPG − 0.005f (Cescin)
For D:TOI ∼ 400 + 0.01Height − 0.005f (Cescin) − 0.03ID IQMJHL
One regression, but three different relationships
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 26 / 35
![Page 78: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/78.jpg)
Model
Interactions: Example
IG , IF , ID Indicator (Positions), IQMJHL Indicator (QMJHL)
Full (Grossly Hypothetical) Model:
TOI ∼ 400 + 60IG + 0.01Height + 0.2PPG × IF + 0.25Height × IG
− 0.005 × f (Cescin) − 0.03ID × IQMJHL
For G:TOI ∼ 400 + 60 + (0.01 + 0.25)Height − 0.005f (Cescin)
For F:TOI ∼ 400 + 0.01Height + 0.2PPG − 0.005f (Cescin)
For D:TOI ∼ 400 + 0.01Height − 0.005f (Cescin) − 0.03ID IQMJHL
One regression, but three different relationships
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 26 / 35
![Page 79: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/79.jpg)
Model
Interactions: Example
IG , IF , ID Indicator (Positions), IQMJHL Indicator (QMJHL)
Full (Grossly Hypothetical) Model:
TOI ∼ 400 + 60IG + 0.01Height + 0.2PPG × IF + 0.25Height × IG
− 0.005 × f (Cescin) − 0.03ID × IQMJHL
For G:TOI ∼ 400 + 60 + (0.01 + 0.25)Height − 0.005f (Cescin)
For F:TOI ∼ 400 + 0.01Height + 0.2PPG − 0.005f (Cescin)
For D:TOI ∼ 400 + 0.01Height − 0.005f (Cescin) − 0.03ID IQMJHL
One regression, but three different relationships
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 26 / 35
![Page 80: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/80.jpg)
Model
Interactions: Example
IG , IF , ID Indicator (Positions), IQMJHL Indicator (QMJHL)
Full (Grossly Hypothetical) Model:
TOI ∼ 400 + 60IG + 0.01Height + 0.2PPG × IF + 0.25Height × IG
− 0.005 × f (Cescin) − 0.03ID × IQMJHL
For G:TOI ∼ 400 + 60 + (0.01 + 0.25)Height − 0.005f (Cescin)
For F:
TOI ∼ 400 + 0.01Height + 0.2PPG − 0.005f (Cescin)
For D:TOI ∼ 400 + 0.01Height − 0.005f (Cescin) − 0.03ID IQMJHL
One regression, but three different relationships
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 26 / 35
![Page 81: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/81.jpg)
Model
Interactions: Example
IG , IF , ID Indicator (Positions), IQMJHL Indicator (QMJHL)
Full (Grossly Hypothetical) Model:
TOI ∼ 400 + 60IG + 0.01Height + 0.2PPG × IF + 0.25Height × IG
− 0.005 × f (Cescin) − 0.03ID × IQMJHL
For G:TOI ∼ 400 + 60 + (0.01 + 0.25)Height − 0.005f (Cescin)
For F:TOI ∼ 400 + 0.01Height + 0.2PPG − 0.005f (Cescin)
For D:TOI ∼ 400 + 0.01Height − 0.005f (Cescin) − 0.03ID IQMJHL
One regression, but three different relationships
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 26 / 35
![Page 82: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/82.jpg)
Model
Interactions: Example
IG , IF , ID Indicator (Positions), IQMJHL Indicator (QMJHL)
Full (Grossly Hypothetical) Model:
TOI ∼ 400 + 60IG + 0.01Height + 0.2PPG × IF + 0.25Height × IG
− 0.005 × f (Cescin) − 0.03ID × IQMJHL
For G:TOI ∼ 400 + 60 + (0.01 + 0.25)Height − 0.005f (Cescin)
For F:TOI ∼ 400 + 0.01Height + 0.2PPG − 0.005f (Cescin)
For D:TOI ∼ 400 + 0.01Height − 0.005f (Cescin) − 0.03ID IQMJHL
One regression, but three different relationshipsc©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 26 / 35
![Page 83: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/83.jpg)
Results
Model Evaluation
Use model to predict TOI (or GP) for each out of sample player incorpus. Rank order players from those.
Calculate Spearman’s Rank Correlation for:
NHL Draft Order vs Actual TOI Order
Predicted TOI Order vs Actual TOI Order
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 27 / 35
![Page 84: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/84.jpg)
Results
Model Evaluation
Use model to predict TOI (or GP) for each out of sample player incorpus. Rank order players from those.
Calculate Spearman’s Rank Correlation for:
NHL Draft Order vs Actual TOI Order
Predicted TOI Order vs Actual TOI Order
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 27 / 35
![Page 85: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/85.jpg)
Results
Results: Corpus Drafted Players
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 28 / 35
![Page 86: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/86.jpg)
Results
Results: Corpus Drafted or CSS Ranked Players
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 29 / 35
![Page 87: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/87.jpg)
Results
Prediction 2016 NHL Draft
Source: Shinzawa 2018 Article
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 30 / 35
![Page 88: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/88.jpg)
Results
Prediction 2016 NHL Draft
Source: Shinzawa 2018 Article
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 31 / 35
![Page 89: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/89.jpg)
Results
Comments
Corpus (which players are included) is important (Cf Table 4 & 5)
This model is not great, lots of ways to improve
Add age and birthdate into modelPsychological metricsAdditional Years of Data, draft -1, draft -2, SVPctAdd Combine Data, Internal Scouting EvaluationTeam quality metricsDifferent methods CART, BART, etc.Gamma link in GAMMore Leagues, drop Other (Bayesian)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 32 / 35
![Page 90: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/90.jpg)
Results
Comments
Corpus (which players are included) is important (Cf Table 4 & 5)
This model is not great, lots of ways to improve
Add age and birthdate into modelPsychological metricsAdditional Years of Data, draft -1, draft -2, SVPctAdd Combine Data, Internal Scouting EvaluationTeam quality metricsDifferent methods CART, BART, etc.Gamma link in GAMMore Leagues, drop Other (Bayesian)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 32 / 35
![Page 91: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/91.jpg)
Results
Comments
Corpus (which players are included) is important (Cf Table 4 & 5)
This model is not great, lots of ways to improve
Add age and birthdate into model
Psychological metricsAdditional Years of Data, draft -1, draft -2, SVPctAdd Combine Data, Internal Scouting EvaluationTeam quality metricsDifferent methods CART, BART, etc.Gamma link in GAMMore Leagues, drop Other (Bayesian)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 32 / 35
![Page 92: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/92.jpg)
Results
Comments
Corpus (which players are included) is important (Cf Table 4 & 5)
This model is not great, lots of ways to improve
Add age and birthdate into modelPsychological metrics
Additional Years of Data, draft -1, draft -2, SVPctAdd Combine Data, Internal Scouting EvaluationTeam quality metricsDifferent methods CART, BART, etc.Gamma link in GAMMore Leagues, drop Other (Bayesian)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 32 / 35
![Page 93: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/93.jpg)
Results
Comments
Corpus (which players are included) is important (Cf Table 4 & 5)
This model is not great, lots of ways to improve
Add age and birthdate into modelPsychological metricsAdditional Years of Data, draft -1, draft -2, SVPct
Add Combine Data, Internal Scouting EvaluationTeam quality metricsDifferent methods CART, BART, etc.Gamma link in GAMMore Leagues, drop Other (Bayesian)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 32 / 35
![Page 94: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/94.jpg)
Results
Comments
Corpus (which players are included) is important (Cf Table 4 & 5)
This model is not great, lots of ways to improve
Add age and birthdate into modelPsychological metricsAdditional Years of Data, draft -1, draft -2, SVPctAdd Combine Data, Internal Scouting Evaluation
Team quality metricsDifferent methods CART, BART, etc.Gamma link in GAMMore Leagues, drop Other (Bayesian)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 32 / 35
![Page 95: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/95.jpg)
Results
Comments
Corpus (which players are included) is important (Cf Table 4 & 5)
This model is not great, lots of ways to improve
Add age and birthdate into modelPsychological metricsAdditional Years of Data, draft -1, draft -2, SVPctAdd Combine Data, Internal Scouting EvaluationTeam quality metrics
Different methods CART, BART, etc.Gamma link in GAMMore Leagues, drop Other (Bayesian)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 32 / 35
![Page 96: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/96.jpg)
Results
Comments
Corpus (which players are included) is important (Cf Table 4 & 5)
This model is not great, lots of ways to improve
Add age and birthdate into modelPsychological metricsAdditional Years of Data, draft -1, draft -2, SVPctAdd Combine Data, Internal Scouting EvaluationTeam quality metricsDifferent methods CART, BART, etc.
Gamma link in GAMMore Leagues, drop Other (Bayesian)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 32 / 35
![Page 97: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/97.jpg)
Results
Comments
Corpus (which players are included) is important (Cf Table 4 & 5)
This model is not great, lots of ways to improve
Add age and birthdate into modelPsychological metricsAdditional Years of Data, draft -1, draft -2, SVPctAdd Combine Data, Internal Scouting EvaluationTeam quality metricsDifferent methods CART, BART, etc.Gamma link in GAM
More Leagues, drop Other (Bayesian)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 32 / 35
![Page 98: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/98.jpg)
Results
Comments
Corpus (which players are included) is important (Cf Table 4 & 5)
This model is not great, lots of ways to improve
Add age and birthdate into modelPsychological metricsAdditional Years of Data, draft -1, draft -2, SVPctAdd Combine Data, Internal Scouting EvaluationTeam quality metricsDifferent methods CART, BART, etc.Gamma link in GAMMore Leagues, drop Other (Bayesian)
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 32 / 35
![Page 99: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/99.jpg)
Results
Return to Themes
Evaluation of players is hard
and it takes time (players develop at different rates from 18yo)
Metrics for assessment are not simple
nor single agreed response metric
In short, noisy and sparse data with slow feedback and
(to teams) results incredibly important
...and it is possible to improve drafting in the NHL via StatisticsModels/Analytics.“. . . [I]t is a numbers game.”
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 33 / 35
![Page 100: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/100.jpg)
Results
Return to Themes
Evaluation of players is hard
and it takes time (players develop at different rates from 18yo)
Metrics for assessment are not simple
nor single agreed response metric
In short, noisy and sparse data with slow feedback and
(to teams) results incredibly important
...and it is possible to improve drafting in the NHL via StatisticsModels/Analytics.“. . . [I]t is a numbers game.”
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 33 / 35
![Page 101: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/101.jpg)
Updates
CESCIN Update
Work with Stat major, Amanda Butterfield (D), St. Lawrence University’20
Join CSS data & Draft Selection data
Larger Data set (2003-2019)
Tweaks to CESCIN values
Ongoing project to build more complete data, eg. 2003
Link to WriteupLink to Data
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 34 / 35
![Page 102: A Look at Draft By Numbers: Using Data and Analytics to Improve …myslu.stlawu.edu/~msch/Schuckers_HANIC2020_DbN.pdf · 1 EDM: Ryan Nugent-Hopkins (F) 2 COL: Gabriel Landeskog (F)](https://reader034.vdocuments.us/reader034/viewer/2022050219/5f650b669c476f6c0a4557ea/html5/thumbnails/102.jpg)
Updates
The End
Talkon 2019 Review of Analytics-based NHL Draft WorkMy Other Papers in Hockey
c©M. Schuckers (SLU) HANIC-Schuckers May 2, 2020 35 / 35