Jump to content
Leader

Random Packer News & Notes

Recommended Posts

@boratt

Draft position isn't weighted very highly, so taking it out only affects the model a bit. The correlation to AV drops from 0.45 to 0.43 (draft position correlated at -0.31). Here's the results, which end up pretty similar to with draft position included:

2006-2015 WR Model Results (w/o Draft Position)
GROUP: Moderate Risk I (Projected Bust Chance: <50%)
Results: 33.33% (14/42) Busts
Year Name College NFL AV Bust Chance PB Chance
2010 D. Bryant Oklahoma State 11.5 2.74% 22.55%
2010 D. Thomas Georgia Tech 13.5 3.17% 22.44%
2015 A. Cooper Alabama 4 6.77% 21.53%
2013 S. Bailey West Virginia 0.5 12.71% 20.04%
2012 K. Wright Baylor 4 20.13% 18.17%
2014 D. Adams Fresno State 9 21.19% 17.90%
2007 C. Johnson Georgia Tech 8.5 26.07% 16.68%
2014 J. Matthews Vanderbilt 5 29.88% 15.72%
2015 T. Lockett Kansas State 8 29.88% 15.72%
2012 J. Blackmon Oklahoma State 0 30.73% 15.50%
2014 B. Cooks Oregon State 12 31.79% 15.24%
2013 T. Williams Baylor 5.5 32.64% 15.02%
2009 H. Nicks North Carolina 9 32.85% 14.97%
2013 D. Hopkins Clemson 9 33.70% 14.76%
2010 J. Shipley Texas 1 33.70% 14.76%
2010 G. Tate Notre Dame 8.5 35.39% 14.33%
2014 J. Landry LSU 8.5 35.82% 14.22%
2006 G. Jennings Western Michigan 11.5 36.67% 14.01%
2008 J. Nelson Kansas State 10 37.94% 13.69%
2015 N. Agholor USC 7 38.36% 13.58%
2014 S. Watkins Clemson 5 39.64% 13.26%
2012 A.J. Jenkins Illinois 0.5 40.06% 13.16%
2007 J. Jones San Jose State 5 40.27% 13.10%
2008 J. Hardy Indiana 0 41.33% 12.84%
2013 T. Austin West Virginia 5.5 44.09% 12.14%
2009 J. Maclin Missouri 6.5 45.15% 11.88%
2015 K. White West Virginia 0.5 45.36% 11.82%
2007 M. Sims-Walker UCF 7 46.84% 11.45%
2012 R. Randle LSU 7.5 47.06% 11.40%
2012 M. Sanu Rutgers 6.5 47.06% 11.40%
2009 M. Crabtree Texas Tech 9.5 47.48% 11.29%
2007 R. Meachem Tennessee 7 47.69% 11.24%
2007 J. Lee Higgins UTEP 1.5 47.69% 11.24%
2007 D. Bowe LSU 8 47.90% 11.18%
2009 K. Britt Rutgers 3 48.33% 11.08%
2008 D. Avery Houston 0 48.33% 11.08%
2014 A. Robinson Penn State 3 48.54% 11.02%
2009 P. Harvin Florida 10.5 48.96% 10.92%
2008 D. Thomas Michigan State 0 48.96% 10.92%
2011 T. Smith Maryland 7.5 49.18% 10.86%
2011 L. Hankerson Miami 1 49.39% 10.81%
2008 M. Manningham Michigan 6 49.81% 10.70%
GROUP: Moderate Risk II (Projected Bust Chance: 50-60%)
Results: 56.00% (14/25) Busts
2011 J. Jones Alabama 9.5 50.02% 10.65%
2012 M. Floyd Notre Dame 7 50.02% 10.65%
2014 K. Benjamin Florida State 6.5 50.45% 10.54%
2006 S. Holmes Ohio State 9 50.87% 10.44%
2014 O. Beckham Jr. LSU 6 51.08% 10.38%
2012 S. Hill Georgia Tech 0 51.08% 10.38%
2014 P. Richardson Colorado 4 51.08% 10.38%
2007 S. Rice South Carolina 8 51.30% 10.33%
2014 J. Huff Oregon 1 51.30% 10.33%
2013 M. Wheaton Oregon State 3 51.51% 10.28%
2011 V. Brown San Diego State 3 51.72% 10.22%
2014 M. Evans Texas A&M 9 51.93% 10.17%
2010 E. Sanders SMU 6 52.14% 10.12%
2006 C. Jackson Florida 0 53.42% 9.80%
2015 J. Strong Arizona State 0 53.63% 9.74%
2010 B. LaFell LSU 6 54.26% 9.58%
2015 D. Green-Beckham Missouri 0 54.69% 9.48%
2007 D. Jarrett USC 1 55.75% 9.21%
2011 T. Young Boise State 0 56.17% 9.10%
2011 A. Pettis Boise State 2.5 58.29% 8.57%
2009 J. Iglesias Oklahoma 0 58.29% 8.57%
2006 M. Stovall Notre Dame 1 59.35% 8.30%
2012 T.Y. Hilton FIU 10 59.57% 8.25%
2006 B. Williams Wisconsin 0 59.78% 8.20%
2013 R. Woods USC 5.5 59.99% 8.14%
GROUP: High Risk (Projected Bust Chance: 60-80%)
Results: 75.00% (24/32) Busts
2007 T. Ginn Jr. Ohio State 4 60.20% 8.09%
2007 S. Smith USC 7.5 60.41% 8.04%
2011 A.J. Green Georgia 11 61.26% 7.82%
2010 D. Williams USC 2 61.90% 7.66%
2015 D. Smith Ohio State 0 62.75% 7.45%
2012 R. Broyles Oklahoma 0 63.38% 7.29%
2012 A. Jeffery South Carolina 8 63.59% 7.24%
2014 C. Latimer Indiana 1.5 64.44% 7.02%
2006 D. Hagan Arizona State 0.5 64.65% 6.97%
2009 B. Robiskie Ohio State 0 65.50% 6.76%
2013 J. Hunter Tennessee 2 65.93% 6.65%
2008 E. Bennett Vanderbilt 4 65.93% 6.65%
2015 B. Perriman UCF 1.5 66.14% 6.60%
2009 M. Massaquoi Georgia 2 66.14% 6.60%
2008 M. Kelly Oklahoma 0 66.35% 6.54%
2007 A. Gonzalez Ohio State 0.5 67.20% 6.33%
2009 P. Turner USC 0.5 67.83% 6.17%
2008 A. Caldwell Florida 3 68.68% 5.96%
2008 H. Douglas Louisville 4 70.38% 5.53%
2015 P. Dorsett Miami 2.5 71.65% 5.21%
2015 D. Funchess Michigan 6.5 72.29% 5.05%
2011 J. Baldwin Pittsburgh 0 72.71% 4.94%
2011 J. Jernigan Troy 1 72.92% 4.89%
2015 C. Conley Georgia 3 73.35% 4.78%
2009 M. Wallace Mississippi 8.5 73.98% 4.62%
2010 T. Price Ohio State 0 74.20% 4.57%
2007 C. Davis LSU 2 74.62% 4.46%
2009 D. Butler Penn State 0 74.62% 4.46%
2014 M. Lee USC 6.5 75.26% 4.30%
2008 D. Jackson California 9.5 76.10% 4.09%
2014 D. Moncrief Mississippi 3 76.32% 4.03%
2010 E. Decker Minnesota 11.5 78.22% 3.55%
GROUP: Very High Risk (Projected Bust Chance: 80-99.99%)
Results: 90.91% (20/22) Busts
2008 E. Doucet LSU 3.5 82.04% 2.59%
2015 D. Parker Louisville 3.5 82.68% 2.43%
2015 S. Coates Auburn 0 83.52% 2.22%
2006 S. Moss Miami 1 83.74% 2.17%
2009 D. Heyward-Bey Maryland 6 86.07% 1.58%
2013 A. Dobson Marshall 0.5 86.49% 1.47%
2007 J. Hill Washington State 2 86.49% 1.47%
2013 C. Patterson Tennessee 3.5 87.55% 1.21%
2011 R. Cobb Kentucky 10 88.61% <1.00%
2009 D. Williams Penn State 0 89.67% <1.00%
2008 E. Royal Virginia Tech 3.5 92.22% <1.00%
2012 T.J. Graham N.C. State 0.5 92.85% <1.00%
2010 A. Benn Illinois 0 94.13% <1.00%
2006 W. Reid Florida State 0 97.73% <1.00%
2007 Y. Figurs Kansas State 0 >99.90% <1.00%
2013 M. Goodwin Texas 2 >99.90% <1.00%
2006 T. Wilson Oklahoma 0 >99.90% <1.00%
2008 L. Sweed Texas 0 >99.90% <1.00%
2012 D. Posey Ohio State 0 >99.90% <1.00%
2011 G. Little North Carolina 2 >99.90% <1.00%
2007 P. Williams Fresno State 0 >99.90% <1.00%
2009 B. Tate North Carolina 1.5 >99.90% <1.00%

Share this post


Link to post
Share on other sites
1 hour ago, AlNFL19 said:

Do you all think this is a pretty good track record for a statistical model? The results here are my WR projection model's projected bust chance and Pro Bowl chance against their NFL Approximate Value. A "Bust" is defined as <5.0 AV (average of years 3-4) for the purposes of the model, and "Pro Bowl" is defined as 10.0+. The "NFL AV" column is the player's average AV in Years 3-4, with "Busts" in red and "Pro Bowlers" in purple (green = hit, so just not a bust).

Are you using pro football reference CaV to determine the Approximate value?   

 

1 of the biggest indicators of NFL WR "success" that I have seen referenced is the % of the college passing yardage/production a WR has, and the sooner that happens in the college career, the better.  

Breakout age and % college production

Share this post


Link to post
Share on other sites
1 minute ago, squire12 said:

Are you using pro football reference CaV to determine the Approximate value?   

 

1 of the biggest indicators of NFL WR "success" that I have seen referenced is the % of the college passing yardage/production a WR has, and the sooner that happens in the college career, the better.  

Breakout age and % college production

Yes, it uses the Pro Football Reference statistic so that I can be consistent across every position (the goal is to have a model for everything eventually), and because it's a pretty good metric all things considered.

Share this post


Link to post
Share on other sites
6 minutes ago, AlNFL19 said:

Yes, it uses the Pro Football Reference statistic so that I can be consistent across every position (the goal is to have a model for everything eventually), and because it's a pretty good metric all things considered.

Need to consider that the calculation of career value is a descending representation of a players best season 100, 90, 80, 70, etc.  So the career value of a player that has played 10 years will be better as a raw number, and the average is affected by the length of the career.  Comparing players that are only in the NFL for 3-4 years against long established veterans gets tricky.  

Snaps played, targets/opportunities will affect a players hit/bust in some manner.  A player like Jordy Nelson might have been considered a bust after 3-4 years in part because the stacked WR position that GB had in front of him (Driver, Jennings, Jones)

Share this post


Link to post
Share on other sites
Posted (edited)
12 minutes ago, squire12 said:

Need to consider that the calculation of career value is a descending representation of a players best season 100, 90, 80, 70, etc.  So the career value of a player that has played 10 years will be better as a raw number, and the average is affected by the length of the career.  Comparing players that are only in the NFL for 3-4 years against long established veterans gets tricky.  

Snaps played, targets/opportunities will affect a players hit/bust in some manner.  A player like Jordy Nelson might have been considered a bust after 3-4 years in part because the stacked WR position that GB had in front of him (Driver, Jennings, Jones)

You're absolutely right. The reason why this model projects in terms of Years 3-4 AV is because it originated from a draft value study that I was doing. Years 3 and 4 are, usually, around the start of a player's "peak" (Football Outsiders says the same with QBASE and whatnot), and they're both on the first contract for every player. That way, it sort of represents telling a GM a projection of what you're going to be guaranteed to get out of a player without having to negotiate a second contract. That does end up with some numbers that don't look great, but mostly, it does a pretty good job of what it's supposed to do.

Edit: it uses average AV in Years 3-4, not career weighted AV.

Edited by AlNFL19

Share this post


Link to post
Share on other sites
2 hours ago, AlNFL19 said:

Do you all think this is a pretty good track record for a statistical model? The results here are my WR projection model's projected bust chance and Pro Bowl chance against their NFL Approximate Value. A "Bust" is defined as <5.0 AV (average of years 3-4) for the purposes of the model, and "Pro Bowl" is defined as 10.0+. The "NFL AV" column is the player's average AV in Years 3-4, with "Busts" in red and "Pro Bowlers" in purple (green = hit, so just not a bust).

2006-2015 WR Model Results
GROUP: Moderate Risk I (Projected Bust Chance: <50%)
Results: 26.83% (11/41) Busts
Year Name College NFL AV Bust Chance PB Chance
2010 D. Bryant Oklahoma State 11.5 6.56% 21.58%
2010 D. Thomas Georgia Tech 13.5 6.77% 21.53%
2015 A. Cooper Alabama 4 8.04% 21.21%
2012 K. Wright Baylor 4 21.61% 17.80%
2013 S. Bailey West Virginia 0.5 22.67% 17.53%
2007 C. Johnson Georgia Tech 8.5 24.79% 17.00%
2014 D. Adams Fresno State 9 25.85% 16.73%
2012 J. Blackmon Oklahoma State 0 29.25% 15.88%
2014 B. Cooks Oregon State 12 31.79% 15.24%
2014 J. Matthews Vanderbilt 5 32.43% 15.08%
2009 H. Nicks North Carolina 9 33.70% 14.76%
2013 D. Hopkins Clemson 9 34.12% 14.65%
2015 T. Lockett Kansas State 8 35.18% 14.38%
2014 S. Watkins Clemson 5 37.09% 13.90%
2015 N. Agholor USC 7 37.52% 13.80%
2013 T. Williams Baylor 5.5 38.15% 13.64%
2008 J. Nelson Kansas State 10 38.79% 13.48%
2010 G. Tate Notre Dame 8.5 39.21% 13.37%
2006 G. Jennings Western Michigan 11.5 39.42% 13.32%
2014 J. Landry LSU 8.5 39.85% 13.21%
2012 A.J. Jenkins Illinois 0.5 40.06% 13.16%
2010 J. Shipley Texas 1 40.27% 13.10%
2013 T. Austin West Virginia 5.5 41.12% 12.89%
2008 J. Hardy Indiana 0 42.39% 12.57%
2015 K. White West Virginia 0.5 42.39% 12.57%
2009 J. Maclin Missouri 6.5 43.45% 12.30%
2009 M. Crabtree Texas Tech 9.5 44.51% 12.04%
2007 J. Jones San Jose State 5 45.36% 11.82%
2007 D. Bowe LSU 8 46.21% 11.61%
2011 J. Jones Alabama 9.5 46.21% 11.61%
2007 R. Meachem Tennessee 7 46.42% 11.56%
2012 M. Floyd Notre Dame 7 47.06% 11.40%
2009 P. Harvin Florida 10.5 47.06% 11.40%
2009 K. Britt Rutgers 3 47.48% 11.29%
2008 D. Avery Houston 0 47.69% 11.24%
2014 O. Beckham Jr. LSU 6 47.90% 11.18%
2014 M. Evans Texas A&M 9 48.12% 11.13%
2008 D. Thomas Michigan State 0 48.33% 11.08%
2014 K. Benjamin Florida State 6.5 48.96% 10.92%
2006 S. Holmes Ohio State 9 49.18% 10.86%
2012 R. Randle LSU 7.5 49.81% 10.70%
GROUP: Moderate Risk II (Projected Bust Chance: 50-60%)
Results: 60.87% (14/23) Busts
2014 A. Robinson Penn State 3 50.87% 10.44%
2012 S. Hill Georgia Tech 0 51.08% 10.38%
2011 T. Smith Maryland 7.5 51.08% 10.38%
2007 M. Sims-Walker UCF 7 51.30% 10.33%
2014 P. Richardson Colorado 4 51.30% 10.33%
2007 S. Rice South Carolina 8 51.51% 10.28%
2012 M. Sanu Rutgers 6.5 51.93% 10.17%
2006 C. Jackson Florida 0 52.57% 10.01%
2011 L. Hankerson Miami 1 53.42% 9.80%
2015 D. Green-Beckham Missouri 0 54.05% 9.64%
2007 J. Lee Higgins UTEP 1.5 54.26% 9.58%
2013 M. Wheaton Oregon State 3 55.33% 9.32%
2008 M. Manningham Michigan 6 55.54% 9.26%
2007 T. Ginn Jr. Ohio State 4 55.54% 9.26%
2007 D. Jarrett USC 1 55.54% 9.26%
2011 T. Young Boise State 0 55.75% 9.21%
2014 J. Huff Oregon 1 55.75% 9.21%
2011 V. Brown San Diego State 3 55.96% 9.16%
2011 A.J. Green Georgia 11 55.96% 9.16%
2010 E. Sanders SMU 6 56.17% 9.10%
2015 J. Strong Arizona State 0 56.39% 9.05%
2010 B. LaFell LSU 6 57.66% 8.73%
2013 R. Woods USC 5.5 58.72% 8.46%
GROUP: High Risk (Projected Bust Chance: 60-80%)
Results: 76.32% (29/38) Busts
2007 S. Smith USC 7.5 60.20% 8.09%
2015 D. Smith Ohio State 0 60.84% 7.93%
2011 A. Pettis Boise State 2.5 61.26% 7.82%
2012 A. Jeffery South Carolina 8 62.32% 7.56%
2015 B. Perriman UCF 1.5 62.53% 7.50%
2012 R. Broyles Oklahoma 0 62.96% 7.40%
2009 B. Robiskie Ohio State 0 63.17% 7.34%
2006 B. Williams Wisconsin 0 63.17% 7.34%
2013 J. Hunter Tennessee 2 63.38% 7.29%
2006 M. Stovall Notre Dame 1 63.38% 7.29%
2009 J. Iglesias Oklahoma 0 63.59% 7.24%
2012 T.Y. Hilton FIU 10 63.81% 7.18%
2014 C. Latimer Indiana 1.5 64.23% 7.08%
2010 D. Williams USC 2 64.23% 7.08%
2007 A. Gonzalez Ohio State 0.5 64.23% 7.08%
2009 M. Massaquoi Georgia 2 65.08% 6.86%
2008 M. Kelly Oklahoma 0 65.50% 6.76%
2008 E. Bennett Vanderbilt 4 66.99% 6.38%
2006 D. Hagan Arizona State 0.5 67.20% 6.33%
2015 P. Dorsett Miami 2.5 67.83% 6.17%
2011 J. Baldwin Pittsburgh 0 68.47% 6.01%
2015 D. Funchess Michigan 6.5 69.53% 5.74%
2007 C. Davis LSU 2 70.59% 5.48%
2009 P. Turner USC 0.5 70.59% 5.48%
2014 M. Lee USC 6.5 71.86% 5.16%
2008 H. Douglas Louisville 4 72.50% 5.00%
2008 A. Caldwell Florida 3 72.50% 5.00%
2008 D. Jackson California 9.5 73.77% 4.68%
2015 C. Conley Georgia 3 74.20% 4.57%
2011 J. Jernigan Troy 1 74.62% 4.46%
2009 M. Wallace Mississippi 8.5 75.68% 4.19%
2015 D. Parker Louisville 3.5 75.89% 4.14%
2010 T. Price Ohio State 0 76.53% 3.98%
2009 D. Butler Penn State 0 76.95% 3.87%
2009 D. Heyward-Bey Maryland 6 78.01% 3.61%
2014 D. Moncrief Mississippi 3 78.44% 3.50%
2010 E. Decker Minnesota 11.5 79.71% 3.18%
2006 S. Moss Miami 1 79.92% 3.13%
GROUP: Very High Risk (Projected Bust Chance: 80-99.99%)
Results: 94.74% (18/19) Busts
2013 C. Patterson Tennessee 3.5 81.83% 2.65%
2008 E. Doucet LSU 3.5 82.46% 2.49%
2013 A. Dobson Marshall 0.5 83.95% 2.11%
2015 S. Coates Auburn 0 84.37% 2.01%
2007 J. Hill Washington State 2 85.86% 1.63%
2011 R. Cobb Kentucky 10 86.28% 1.53%
2008 E. Royal Virginia Tech 3.5 87.13% 1.31%
2010 A. Benn Illinois 0 88.61% <1.00%
2009 D. Williams Penn State 0 89.25% <1.00%
2012 T.J. Graham N.C. State 0.5 90.73% <1.00%
2006 W. Reid Florida State 0 97.73% <1.00%
2007 Y. Figurs Kansas State 0 99.64% <1.00%
2013 M. Goodwin Texas 2 >99.90% <1.00%
2008 L. Sweed Texas 0 >99.90% <1.00%
2006 T. Wilson Oklahoma 0 >99.90% <1.00%
2011 G. Little North Carolina 2 >99.90% <1.00%
2012 D. Posey Ohio State 0 >99.90% <1.00%
2007 P. Williams Fresno State 0 >99.90% <1.00%
2009 B. Tate North Carolina 1.5 >99.90% <1.00%

How can something be evaluated if we don't have your metrics?

Share this post


Link to post
Share on other sites
7 minutes ago, AlexGreen#20 said:

How can something be evaluated if we don't have your metrics?

By the results?

Share this post


Link to post
Share on other sites

You shouldn't say your model has a "track record" or "results" unless you have predicted out of sample data.

Share this post


Link to post
Share on other sites
51 minutes ago, AlNFL19 said:
59 minutes ago, AlexGreen#20 said:

How can something be evaluated if we don't have your metrics?

By the results?

So the players that have better results are less likely to bust?

That is some stellar analytics

Share this post


Link to post
Share on other sites
26 minutes ago, squire12 said:

So the players that have better results are less likely to bust?

That is some stellar analytics

Not sure what you’re talking about here. He calculated their bust probability based on college production and other things; then posted the results of that compared to their career results. 

Share this post


Link to post
Share on other sites
39 minutes ago, Julyan Morley said:

You shouldn't say your model has a "track record" or "results" unless you have predicted out of sample data.

He literally posted a sample size of 10+ years.

Share this post


Link to post
Share on other sites

Bros...... are you even sciencing right?

Share this post


Link to post
Share on other sites
2 hours ago, AlNFL19 said:

By the results?

That is not at all how science or mathematics works. 

Post your methodology.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×