In this week's Varsity Numbers, Bill Connelly revisits some measures and concepts: Adjusted Scores, Covariance, and momentum (or whatever you choose to call it).
25 Apr 2012
by Bill Connelly
It's that time of year again. Allergies are exploding, strange storms are popping up around the country, the NFL Draft is this weekend ... it must be time for an Adj. POE column!
POE stands for "Points Over Expected." The idea for POE is simple: It compares a runner's production (in terms of EqPts) to the production that would have been expected of an average back given the same carries against the same opponents. A runner with, say, a plus-6.0 Adj. POE produced the equivalent of a touchdown more value than the average FBS running back would have with the same carries.
POE = EqPts - Expected EqPts.
Last year, we added an adjustment to account for the quality of a runner's offensive line (based on the team's Adj. Line Yards ratings). So below, you will see a raw POE figure and the more comprehensive Adj. POE figure. Backs are ranked by Adj. POE.
Now let's expand on another topic we unveiled in last year's POE column: Highlight Yards.
A few months ago, FO introduced the idea of "Second Level Yards" and "Open Field Yards" for NFL running backs. These were the remaining yards that were left after each break in the baselines for Adjusted Line Yards. I've done the same thing here for college backs, with two differences. First, we're adding together both "Second Level" Yards (5-10 past the line) and "Open Field" Yards (11-plus past the line). Second, we're counting only half the Second Level Yards, just as the line gets half credit for these yards. We'll call this stat "Highlight Yards," because these longer runs are the ones that show up on the highlight shows. A three-yard run gets zero Highlight Yards. A 70-yard run gets 63 Highlight Yards. The more Highlight Yards, the more explosive the runner was, and the less his overall yardage and POE totals were due to the offensive line blocking for him.
Basically, Highlight Yards are the yards credited to the running back and not the blocking. Again, we rank them below in terms of their full-season accumulation instead of their per-carry average.
In a nutshell, Adj. POE compares your performance to that of an average runner, and Highlight Yards measure both your explosiveness and the frequency with which you pull off explosive runs.
Montee Ball's plus-57.2 Adj. POE is the highest of the "play-by-play data" era (2005-11).
Top 10 Single-Season Adj. POE (2005-11)
1. 2011 Montee Ball, Wisconsin (plus-57.2)
2. 2010 Alex Green, Hawaii (plus-55.0)
3. 2008 Jahvid Best, California (plus-51.3)
4. 2008 Colin Kaepernick, Nevada (plus-48.6)
5. 2008 Yonus Davis, San Jose State (plus-45.8)
6. 2011 Trent Richardson, Alabama (plus-42.6)
7. 2006 Pat White, West Virginia (plus-41.2)
8. 2008 Brandon Rutley, San Jose State (plus-40.2)
9. 2008 LeSean McCoy, Pittsburgh (plus-40.2)
10. 2008 Knowshon Moreno, Georgia (plus-38.5)
Memphis' Artaves Gibson almost had the worst rushing season in recent memory.
Worst 10 Single-Season Adj. POE (2005-11)
1. 2008 Chip Bowden, Army (minus-32.7)
2. 2011 Artaves Gibson, Memphis (minus-31.5)
3. 2008 Deonte' Jackson, Idaho (minus-27.0)
4. 2008 Darius Marshall, Marshall (minus-23.6)
5. 2011 Traylon Durham, Kent State (minus-23.5)
6. 2008 Frank Goodin, Louisiana-Monroe (minus-21.3)
7. 2010 MiQuale Lewis, Ball State (minus-20.1)
8. 2009 Seth Smith, New Mexico State (minus-20.0)
9. 2007 Deonte' Jackson, Idaho (minus-19.0)
10. 2009 Thomas Merriweather, Miami (Ohio) (minus-18.8)
The good news for Gibson: it gets better. Maybe. Deonte' Jackson has the dubious honor of making this list twice -- after a minus-19.0 season his freshman year, he fell to minus-27.0 his sophomore year. Then he improved to minus-8.1 in 2009 and minus-8.4 in 2010! So there's hope ... that Gibson may inch close to becoming an average running back. (And yes, there's a reason why Jerrell Rhodes is considered the favorite for the Memphis starting running back job this year.)
The ebb and flow of the yearly Adj. POE totals has been fascinating. There was an enormous spread between good and bad in 2008 (highest: plus-51.3; lowest: minus-32.7) and 2011 (highest: plus-57.2; lowest: minus-31.5). But in other years parity has reigned. The highest Adj. POE in 2005 was just plus-32.9 (Reggie Bush), the highest in 2007 was plus-33.9 (Dan LeFevour), the highest in 2009 was plus-34.9 (the immortal Lance Dunbar of North Texas) and, aside from Outlier Alex Green, the second-highest in 2010 was plus-26.9. The 2011 season was the most ridiculously widespread of all. What that means, I will leave to others to decide.
Each April before the NFL draft, I take a look at how some of Football Outsiders' metrics -- Speed Score, Adj. POE and Highlight Yardage, specifically -- can be used to predict pro success. Through the first couple years, the results have been spotty at best. Last year, I correctly (so far) identified Roy Helu as rather underrated and threw water on the viability of Jacquizz Rodgers (and, for that matter, Shane Vereen ). I also offered a tepid endorsement, at best, of Mark Ingram. Past endorsements of backs like Ryan Mathews and LeGarrette Blount have turned out well.
But according to the three measures above, Stevan Ridley shouldn't have even been considered in the third round, and he CERTAINLY shouldn't have averaged 5.1 yards per carry over 87 totes. Along those same lines, Evan Royster shouldn't have seen the success he did in 2011. Meanwhile, Alex Green should have set the NFL aflame in his rookie year. (Granted, he suffered a knee injury, but he also only garnered three carries for 11 yards when healthy.) It goes without saying that Cedric Peerman has not quite lived up to his plus-2 status.
Still, let's see what the numbers are telling us that we might have otherwise missed.
Here's how we do it:
Here are the pre-draft Highlight Yards, Adj. POE and Speed Score averages for players in both categories:
- Successful: 2.64 Hlt Yds/carry, +12.7 Adj. POE, 108.5 Speed Score
- Not Successful: 1.89 Hlt Yds/carry, +3.6 Adj. POE, 99.9 Speed Score
So, to analyze a sample of almost 100 rushers, we are going to set up a series of arbitrary points.
- +1 point: more than 2.64 Hlt Yds/carry, +12.7 Adj. POE or more, 108.5 speed score or higher
- -1 point: less than 1.89 Hlt Yds/carry, +3.6 Adj. POE or fewer, 99.9 speed score or lower
- 0 points: everything else
Using this broad system, a player can end up with somewhere between 3 and -3 points. Broken into the four tiers below, you can see defined distance from one tier to another.
Looking at some correlation data, Speed Score does more strongly predict success than the other two measures, but the combination of the three works even better. I could probably give Speed Score double-weight in the formulas here, but the simpler, the better for this exercise.
|Plus-2 Backs||Lamar Miller, Miami
David Wilson, Virginia Tech
Robert Turbin, Utah State
Michael Smith, Utah State
Jonas Gray, Notre Dame
Jeff Demps, Florida
|Plus-1 Backs||Trent Richardson, Alabama
LaMichael James, Oregon
Bernard Pierce, Temple
Terrance Ganaway, Baylor
Vick Ballard, Mississippi State
|Plus-0 Backs||Doug Martin, Boise State
Isaiah Pead, Cincinnati
Ronnie Hillman, San Diego State
Cyrus Gray, Texas A&M
|Minus-1 Backs||Davin Meggett, Maryland||-2.5||1.98||100.2||Late|
|Minus-2 Backs||Chris Polk, Washington
Chris Rainey, Florida
Dan Herron, Ohio State
Brandon Bolden, Ole MIss
|Minus-3 Backs||Edwin Baker, Michigan State
Tauren Poole, Tennessee
Darrell Scott, South Florida
Marc Tyler, USC
Lennon Creer, Louisiana Tech
So what can we learn from this?
1. There are no slam dunks. There rarely are, of course. Trent Richardson is close -- his 2.60 Highlight Yards per carry are obviously quite close to a plus-1 in that category. There is a bit of iffiness concerning the fact that he didn't run the 40 at the combine. If we include the reported 4.49 that he ran on his Pro Day, he would have a Speed Score of 112.2, and become a "plus-2" back. He is probably a better prospect than former teammate Ingram was last year. Beyond Richardson, the other two possible first-rounders have some serious question marks. LaMichael James is not quite as fast as one would hope considering his size, and Doug Martin's line is incredibly nondescript. Martin would be a solid third- or fourth-rounder, and he could pass as a second-rounder, but to be a first-rounder, you need to be great at something. Martin is not.
2. Lamar Miller is the most draftable Miami running back in a while. Javarris James (minus-1), Damien Berry (minus-3) and Graig Cooper (minus-3) did not really pass the sniff test, but Miller put together a sturdy stat line. He was 0.09 Highlight Yards per carry away from a plus-3 rating.
3. 2011 was the year of the Utah State running back. Gary Andersen is building a sturdy program at USU, but it probably goes without saying that the backfield took a hit with the loss of TWO professional-caliber backs. Smith's Speed Score is listed as "N/A" because it was not recorded at the NFL Combine, but if the 4.35 he ran at USU's Pro Day is plugged in, that's a Speed Score of 114.5 and a plus-3 rating for him.
4. The pool is very deep this year. Do not waste a pick on Edwin Baker, Tauren Poole or Chris Polk. Polk rushed for 4,014 yards in a Washington uniform, and that's a pretty number, of course, but there are red flags all over the place for him. If this was a thin pool of talent, you could take a flier on these guys, but there are quite a few backs more worthy of a fourth-, fifth-, sixth- or seventh-round pick.
5. Take a chance on Michael Smith, Jonas Gray, Jeff Demps, Terrance Ganaway, or Vick Ballard. Each of these backs have their own red flags despite solid stats. Smith is small (5-foot-9), relatively untested in the WAC, and missed most of 2010 with a foot injury. Gray tore his ACL last year. Demps is small and injury-prone. Ganaway was a bit of a one-year wonder no matter how intriguing his size is. Ballard isn't very fast. But these are reasons why these players aren't being considered in the top three or four rounds. They are still probably more worthy of a pick than the players listed in Note No. 4 above.
6. Your final season is more telling than your career stats. This week I investigated whether career stats should be used in some regard, instead of just last season's numbers. It would make sense -- in 2009, Jacquizz Rodgers was an Adj. POE marvel (plus-26.9), and in 2010 he was mediocre (plus-1.7); doesn't 2009 say something about his potential? But no matter what I tried, the correlations got weaker when I used previous years' data, even with a lower weight. I will continue to explore, but for now, we will stick with just last year's data.
5 comments, Last at 22 Feb 2013, 4:18pm by dinelson