Thursday, December 12, 2013

Reprint of QB selection and the NFL draft

Richard Thaler is likely very happy that RGIII is not doing well in Washington.  While the Redskins may very well have overpaid for RGIII I believe that that the evidence supporting the winner's curse is substantially overstated.

Most later-round QB picks don't stay in the league long.  Tom Brady was an aberration.  Had he been picked by Cleveland rather than New England he probably would not have been noticed.

Some time next week I will post some material on winner's curse.   Today I am reprinting this post on QB selection in draft and subsequent QB performance.


Question:  To what extent is the order of QB selection in the draft associated with subsequent QB career performance?

Answer:  The most often cited measure of QB performance used in the NFL is a composite QB rating index.  The composite index incorporates information on four factors,  -- completions as a percent of attempts, yards per attempt, touchdowns per attempt, and interceptions per attempt.   (A description of the NFL QB rating index can be found in Appendix One.) 

When comparing QB performance based on a performance statistic like the NFL rating variable or an interception percentage it makes sense to exclude QBs with relatively few career pass attempts for two reasons.  First, the statistic for a QB with relatively few pass attempts might be affected by a small number of extreme performances in a relatively small number of games.  Second, the inclusion of few-pass-attempt QBs in performance comparisons is misleading because the opportunity cost of selecting a QB that does not get much playing time and spends only a short period of time in the league is limited to the lost value of a draft pick rather than an endless stream of bad games.  In essence, management mitigates the cost of a bad draft pick by cutting the player from the team.

Statistics below depict differences in NFL QB rating across QB-choice groups for all QBs with more the 500 career passing attempts. 

·      The mean QB rating  (75.6) is highest for the first QB chosen but the fourth QB chosen has a higher average QB rating than the second or third QB chosen.  The second QB choice level has the lowest average QB rating (67.0).

·      The performance-gap between the first and second QB choice is large across the distribution.  Historically, the second-choice QB has underperformed.

·      The 25th percentile for the fourth QB chosen is higher than the 25th percentile for all other QB choice groups.

The relatively high QB performance level for the fourth QB chosen is largely the result of the fact that marginal or weaker late-choice QBs are more likely to be cut quickly than marginal early-choice QBs.  Observe from Table Three that 25 of 31 first-choice QBs have more than 500 career passing attempts compared to 13 of 31 fourth-choice QBs. 

There are at least three reasons why a marginal early-choice QB might stay in the league longer and get more playing time than a marginal late-cost player.  First, the marginal early-choice QB may actually be a better QB than the marginal late-choice QB.  Second, management may pay more attention to the early-choice players than the late-choice players.  Third, management might choose to delay cutting a below average early-choice QB in order to avoid acknowledging a mistake.



  QB Rating Statistics for QBs with more than
500 career passing attempts
QB choice order
Mean
25th percentile
50th percentile
75th percentile
# of QBs with more than 500 attempts
1
75.6
69.4
77.1
81.1
25.0
2
67.0
61.9
68.4
72.9
22.0
3
71.6
67.0
71.8
77.8
15.0
4
74.7
70.7
72.3
79.7
13.0






The quarterback comparisons based on pick-choice order involve two issues:  (1) what is the probability of getting a competent quarterback and (2) what is the probability of getting a superstar quarterback.  Additional insight on how these two performance outcomes vary across QB choice number is presented below.

·      The number of first-choice QBs with a QB rating over 85 was equal to the number of fourth-choice QBs with a rating over 85.   The first-choice QBs with a rating over 85 were Peyton Manning and Chad Pennington.  The fourth-choice QBs with a QB rating over 85 were Duante Culpepper and Joe Montana.  There was no second-choice QB with a QB rating over 85. There was only one third-choice QB (Brett Favre) with a rating over 85.

The main difference between the first-QB choice and subsequent QB choices is the number of QBs with a QB rating of 75 or greater.

·      For the 31 draft years covered in this sample, 16 first QB choices had a QB rating 75 or greater, more than the other three rounds combined.  There were 4 second-choice QBs, 5 third-choice QBs, and 6 fourth-choice QBs with a rating over 75.


QB Ratings Frequency Distribution
pick #
1
2
3
4
<55
7
9
11
16
55-64.9
1
7
6
3
65-74.9
7
11
9
6
75-84.9
14
4
4
4
85<=
2
0
1
2
31
31
31
31




No comments:

Post a Comment