As a Cleveland Browns fan, Ohio University adjunct professor William Young knows all about NFL Draft busts. From Mike Junkin to Tim Couch to Gerard Warren, the Browns’ history of draft busts rivals any team, except probably the Detroit Lions, whose fans have to be applauded for not shooting themselves in the face at the end of every season.
A few years ago Young had a brainstorm. Wouldn’t it be great, he thought, if there was a way to apply all of the skills that I have learned through my courses at OU to develop a tool for the Browns to help them with their decisions?
Young has since invented a statistical model to help NFL teams make smarter draft picks. He created the system for his doctoral dissertation and hopes to license it with a software company and/or NFL team.
Young shared his trove of data with Made Man. It’s an NFL Draft junkie’s delight. And someday it might be used to help NFL teams draft more keepers, especially deep in the draft.
UPDATE: Part 4 of our draft package, Made Man’s All-Undrafted NFL Team
The Statistical Model: A Brief Overview
Young developed an artificial neural network – also known as artificial intelligence – that forecasts which players would be best for an individual team to pick during the draft or in free agency. Teams would use the Heuristic Evaluation of Artificially Replaced Teammates, or H.E.A.R.T., to develop a point system that would rank players. The end values represent a player’s ability to affect the expected number of games won in a season if the team added that player to its roster.
“I see my system as a supplemental tool that NFL teams and their scouts can use to maximize team efficiency,” Young says. “They all have limited budgets and only so many players they can hire. My tool helps them choose the right ones at the right price.”
The system is not perfect. Young says it needs more data as well as a layer of expertise applied to the data by NFL executives and coaches.
Still, it’s a start.
A detailed account of how Young developed his system and determined values can be found here.
The NFL Draft Is Important
Rookies are extremely important, according to the model. Out of the top 10 players identified by Young in the 2007 draft and free agent class, 70 percent were rookies. In terms of impact, the top seven rookies had much higher expected contribution values than any free agent.
Last, First … Position … Type … Expected Value (added wins)
Willis, Patrick … LB … Draft … 2.30
Peterson, Adrian … RB … Draft … 1.81
Quinn, Brady … QB … Draft … 1.24
Russell, JaMarcus … QB … Draft … 0.90
Hall, Leon … CB … Draft … 0.88
Carriker, Adam … DT … Draft … 0.85
Ross, Aaron … CB … Draft … 0.60
Steinbach, Eric … OG … Free Agent … 0.53
Thomas, Adalius … LB … Free Agent … 0.53
Porter, Joey … LB … Free Agent … 0.51
The system is not perfect. Obviously, Broncos quarterback Brady Quinn has not proved as valuable as Browns guard Eric Steinbach. But this system also said that JaMarcus Russell should not have been the first quarterback or player taken in the draft. In fact, it’s fair to say that Quinn is a better quarterback, because at least he is still in the league and Russell is not.
To the system’s credit, Willis won defensive rookie of the year and is considered the best linebacker in the game and Peterson was offensive rookie of the year and is probably the NFL’s best running back. Even though he did not make the top 10, H.E.A.R.T. had Jets All-Pro cornerback Darrelle Revis rated highly.
Most Important Attributes By Position
Young found that certain attributes – comprised of NFL Combine numbers and college stats – helped determine whether or not a college player was successful in the pros. Here is the breakdown of most important attributes by position. Made Man has also included the 2011 leaders in each category, when that information was available.
Quarterback: Broad Jump
2011 Combine Leaders: T- Cam Newton (Auburn), Tyrod Taylor (Virginia Tech), Josh Portis (California, Pa.)
Running Back: Receiving Targets
Keep an Eye On: DeMarco Murray (Okahoma) caught 71 passes for 594 yards and 5 touchdowns in 2010
Tight End: 10-Yard Dash
2011 Combine Leader: Rob Housler (Florida Atlantic)
Wide Receiver: Broad Jump
2011 Combine Leader: Julio Jones (Alabama)
Defensive Tackle: Defensive Assists
Defensive End: T- Defensive Stuffs and Broad Jump
2011 Combine Leader (Broad Jump): Ryan Kerrigan (Purdue)
Linebacker: QB Knock Downs
Keep an Eye On: Von Miller (Texas A&M) who had 10.5 sacks in 13 games
Cornerback: 40-Yard Dash
2011 Combine Leader: DeMarcus Van Dyke (Miami of Fla.)
Keep an Eye On: Rahim Moore (UCLA), 14 career interceptions
Note: The attributes for center, guard and offensive tackle were not significant. Offensive line is, according to Young, a difficult position to gauge statistically.
Attributes That Were Not Important By Position
Young’s research also revealed that much of what happens at the NFL Combine provides teams and scouts with useless data. Combine drills include 40-yard dash, bench reps, cone drills and so on. Here is a list of unimportant attributes by position.
Quarterback: Cone Drill
2011 Combine Leader: Jake Locker (Washington)
Running Back: 40-Yard Dash
2011 Combine Leader: Da’Rel Scott (Maryland)
Tight End: Receiving Touchdowns
Wide Receiver: Bench Reps
2011 Combine Leader: Greg Little (North Carolina)
Center: Bench Reps
2011 Combine Leader: Ryan Bartholomew (Syracuse)
Guard: 40-Yard Dash
2011 Combine Leader: Derek Newton (Arkansas State)
Offensive Tackle: Bench Reps
2011 Combine Leader: Chris Hairston (Clemson)
Defensive Tackle: Cone Drill
2011 Combine Leader: Nick Fairley (Auburn)
Defensive End: Bench Reps
2011 Combine Leader: J.J. Watt (Wisconsin)
Linebacker: Shuttle Drill
2011 Combine Leader (60-Yard): Von Miller (Texas A&M)
Safety: Broad Jump
2011 Combine Leader: Aaron Williams (Texas)
The 2011 NFL Draft
So who should Carolina select with the No. 1 pick? Young says no comment. Probably because he’s a Browns fan and he’s hoping for a value pick with the No. 6 selection.