I am ready for some football
September 4, 2009
While I have always enjoyed watching collegiate football, I have never followed professional football with much interest. My typical excuse: some “purity” is lost when athletes turn pro.
Seriously? Waa waa waa waaaaaaaa. Reflecting on this pathetic excuse, I am going to do a few things differently this season.
First, I am not going to use the words “purity” and “football” in the same sentence. Starting now.
Third, I am going to play fantasy football. While this will be a great opportunity to learn more about the various NFL teams and to shoot the breeze about Drew Brees (ugh, did I write that?), I am excited about fantasy football for a different reason: analytics. Indeed, the entire purpose of this series of blog posts is to publicize my attempts to beat the snot out of my opponents with analytics.
So in preparation for my first fantasy football draft, I did what any self-respecting statistician would do; I obtained some empirical data. In particular, I obtained the lifetime statistics for all offensive players—quarterbacks (QBs), running backs (RBs), wide receivers (WRs), tight ends (TEs), and kickers (Ks)—that are playing in the NFL this season. For the purposes of this post, I did not analyze fantasy defenses as this would have required considerably more work.
By combining these game-by-game statistics for each player with the fantasy points allocated in my Yahoo! Sports Fantasy League (i.e. 6 fantasy points per touchdown, 1 fantasy point per 10 yards rushing, etc.), I can compute the distribution of fantasy points for each player. This is somewhat complicated by the fact that some players (er hem, Bret Favre) have been playing in the NFL for a long time. Since it is unlikely that a player's performance 10+ years ago will bear any resemblence to his performance this season, I compute the distribution of fantasy points by using a linear weighting function that places the most weight on games from last season and no weight on any games from 4+ years ago. This weighted distribution of fantasy points therefore places the most emphasis on outcomes from recent seasons without completely neglecting past performances.
I use this weighted fantasy point distribution for each player to determine which player is more likely to score more fantasy points in a game. For example, I rank Drew Brees higher than Adrian Peterson, because Drew Brees is more likely to score more fantasy points in a game than Adrian Peterson (Fig. 1). The results of my rankings compared with the Yahoo! Sports and CBS Sportsline rankings are shown in Tbl. 1.
|My rank||Yahoo! Sports rank||CBS Sports rank|
|1.||Drew Brees (QB)||Adrian Peterson (RB)||Adrian Peterson (RB)|
|2.||Kurt Warner (QB)||Michael Turner (RB)||Maurice Jones-Drew (RB)|
|3.||Peyton Manning (QB)||Maurice Jones-Drew (RB)||Matt Forte (RB)|
|4.||Philip Rivers (QB)||Matt Forte (RB)||Michael Turner (RB)|
|5.||Jay Cutler (QB)||Brian Westbrook (RB)||DeAngelo Williams (RB)|
|6.||Aaron Rodgers (QB)||DeAngelo Williams (RB)||Frank Gore (RB)|
|7.||Donovan McNabb (QB)||LaDainian Tomlinson (RB)||Steve Slaton (RB)|
|8.||Larry Fitzgerald (WR)||Larry Fitzgerald (WR)||Chris Johnson (RB)|
|9.||Deangelo Williams (RB)||Steven Jackson (RB)||Steven Jackson (RB)|
|10.||Tony Romo (QB)||Drew Brees (QB)||Larry Fitzgerald (WR)|
This analysis highlights a few interesting issues. First, I systematically rank quarterbacks higher than running backs and wide receivers whereas Yahoo! Sports and CBS Sportsline appear to do the exact opposite (Tbl. 1). It is unclear to me why Yahoo! Sports and CBS Sportsline rank running backs and wide receivers higher than quarterbacks; quarterbacks clearly have a higher average fantasy points than running backs and wide receivers (Fig. 2). One possibility for this discrepency is that Yahoo! Sports and CBS Sportsline use a much more sophisticated ranking scheme than my minimalistic approach. For instance, they might account for a team's schedule, the effect of Terrell Owens on his teammates, or the quality of a particular player relative to other players of the same position.
Second, it is clear that the distribution of fantasy points does not place equal emphasis on each position (Fig. 2). Quarterbacks systematically have more fantasy points than nearly every other position. In comparison with the other offensive players, tight ends are almost inconsequential additions to any fantasy team. It is interesting to imagine changing the fantasy point allocation such that each position has the same expected distribution of fantasy points. For example, tight ends should receive 1 fantasy point for every 3 yards receiving to put them on par with wide receivers that receive 1 fantasy point for every 10 yards receiving.
Feeling like a broker trading on insider information, I showed up to my first online draft with my precomputed list of ranked players (Tbl. 1). I felt smuggly confident as I drafted all of my offensive players and populated my bench with, dare I say, a superb B squad. When it came to defenses, there was really only one reasonable choice I could make: Da Bears. So now I enter the fantasy football season with my team (Tbl. 2), aptly named “Does not compute”, on a quest to beat the snot out of my opponents.
|QB||Kurt Warner||Jay Cutler|
|RB||Thomas Jones||Steve Slayton||Clinton Portis||Kevin Smith||Tim Hightower|
|WR||Larry Fitzgerald||Brandon Marshall||Santonio Holmes||Vincent Jackson||Hines Ward|