Draft Science, Part I
By Appaloosa
September 2005
About a month before the 2005 NFL Draft I read an article by Len Pasquerelli of ESPN.com in which he stated that the position drafted in the first round most likely to be a bust was defensive end. I had my doubts about that, and it got me to thinking – what is the position most likely to produce a first round bust?
It being deep into the off-season, and having nothing better to do, I decided to conduct a scientific study of first round draft picks by position to determine if any particular position is more prone to produce busts than any other. I was also interested in the general quality of each position. Because it takes a minimum of five years for a player’s quality to be fully recognized (especially separating pro bowlers from future hall of famers and busts from disappointments), I did not assess any drafts more recent than 2000. To give myself sufficient data for a reliable statistical analysis, I needed data on first round draft picks for six drafts. The following study was conducted on the first round draft picks from 1995 through 2000. Each player was categorized according to his position listed at the time he was drafted – not necessarily the one he eventually played (such as a player drafted as a DE who ended up playing DT or a CB who was switched to safety). Because a data set with only one member was useless for this study, I threw out Janikowski, the only kicker selected in the first round during the period in question. Everyone else was grouped into one of 11 categories: QB, RB, WR, TE, OT, center/guard (C/G), DE, DT, LB, safety (S), or CB.
I then developed a grading system to quantify the quality of each player. While the grading was a little subjective, I did my best to rely mainly on statistics specific to the position played, number of starts divided by possible starts, and number of Pro Bowls. Super Bowl rings did not enter into the equation because there are often bad players on good teams (for example Marcus Nash who has 2 Super Bowl rings but was a complete bust). Eventually each player was assigned one of the following grades:
1 – Bust – qualifications include bad stats, few games started, out of the league in less than five years – example Ryan Leaf
2 – Disappointment – these are players who weren’t truly awful, but never quite lived up to expectations – example Tim Couch
3 – Solid Starter – the very least one should expect to get out of a first round draft pick is a guy who will start for your team (or someone else’s) for a minimum of five years and produce respectable stats and might even make the Pro Bowl one year – example Kerry Collins
4 – Pro Bowler – these players have very good careers that span quite a few years, really produce for their team and get voted to the Pro Bowl on one or more occasions – example Steve McNair
5 – Future Hall of Famer – these are the rare players who stand among the best at their position for an extended period of time. Because it appears to take a minimum of six years as a starter for a future Hall of Famer to reveal himself, there are no 5s assigned to anyone drafted in the first round after 1998 though there are a few 1999 picks on the waiting list.
Once all the grades had been assigned, I conducted a bit of statistical analysis on the results. First I computed the mean (average) grade for each group. The results ranged from 2.3 for RBs to 3.3 for TEs. The full table of mean grades is as follows:
QB – 2.6, RB – 2.3, WR – 2.9, OT – 2.9, C/G – 3.1, TE – 3.3, DT – 2.7, DE – 2.8, LB – 3.1, S – 3.0, CB – 2.8
As can be seen above, the mean grade for DEs is 2.8, which is slightly better than the grade for DTs and only slightly worse than the grade for OTs – hardly a guarantee of a bust. So is there really any difference between a mean grade of 2.3 and 3.3?
To determine if there is a significant difference in the grades between the 11 different categories, I conducted an analysis of variance (ANOVA). I won’t go into details except to say that ANOVA is a commonly used statistical test used to determine if there is a difference between one group and an assortment of related groups. The complete results and calculations are available upon request to anyone insane enough to want to see them. The ANOVA equations calculate a “F” value which is then compared to a table “F” value based on the size of the sample. If the calculated “F” value is less than the table value, there is no significant difference between the groups. If the calculated value is greater than the table value, then one or more of the groups is different. Then it is time to look back at the mean grades.
My calculated “F” value for the above data was 0.88, which corresponded to a table value of 1.87. Based on this result, no one position is more likely to result in a first round draft bust than any other.
The things we do to keep ourselves amused during the off-season
Loading ...