A Statistical Look at Draft Value
Over the last couple of days I've been looking at comparing on field production to draft value, and I've come up with some interesting numbers I think.
I used data from
Pro Football Reference to for the five years between 2014 and 2018 to examine the relationship between draft position and their
Career Approximate Value metric. No metric is perfect, but this one generally seems to hold, and it has the benefit of being used across all positions. Stopping at 2018 allows a fair assessment of players drafted in that year, since they have had the opportunity for 3 years of production.
Couldn't get the table to post in a readable way, so
here's the link
I split the data across positions, and across early (1-3) and late (4+) rounds. What I'm trying to do by that is identify how much production varies across early and late round players at different positions, and use the correlation number to identify how predictably a player produces on the field relative to draft position.
In statistics, a correlation coefficient will vary between 1 and -1, with this data, if production were perfectly predictable (ie players drafted higher will always be better), the coefficient would be -1, meaning that a lower draft position number (#1 overall, vs #150 in the 4th round) will always have a higher CareerAV number. The stronger that correlation number, the more value is skewed towards high draft picks, the the inverse also true.
Some predictable results (to me at least):
- QB production is highly predictable. There is a huge difference between the CareerAV of early and late round players and the correlation between draft position and production is the strongest of any position.
- OT production is also highly predictable. Early round players produce, late round players here generally don't. Slightly interesting that from the 4th round on, production is random. Most will not produce, but a guy in the 7th is just as likely to work out as anyone.
- Interior OL production has the weakest relationship between draft position and production. This makes the position poor value early and good value late. Production is also random from R4 onwards. Even isolating players taken within the first 3 rounds, the correlation between draft position and production is not strong, meaning earlier picks in this position are riskier than at other positions.
Some surprising results:
- RB production is relatively predictable. The data says that the NFL is generally effective in splitting players to draft early vs. late, but within R1-3 the correlation is moderate. So, 2nd / 3rd round picks are good value.
- The split in production between early and late round WRs is huge. Pick your WRs early, but within those early rounds, production is not strongly predictable. So, the same as RBs, draft your WRs in the 2nd and 3rd.
- TE production is somewhat predictable, but in TEs drafted in the early rounds, production isn't all that predictable. This is another position where you get get production drafting in early rounds, but within that, the later the better for value.
- LB production is more predictable than average, and there is a solid difference in production between early and late round drafted players.
- All DL players have among the lower difference in production between early and late rounds. I expected this to be true for DT, but not DE. Production is more predictable for DE but value can be had in later rounds at both positions.
- CB production is only moderately predictable and there is less difference between early / late round production than at most positions. I expected the opposite to be true.
- S production is generally predictable, and is predictable even within players drafted in early rounds. In other words, don't be shy about drafting S early.
Some overall conclusions I'm taking from this (ignoring position value in scheme etc):
- Players projected early at QB, OT, S are generally most likely to produce.
- There's a heap of value and flexibility in R2-3, with WR, RB, TE, LB, S picked here, generally likely to produce.
- Stop taking late round flyers on WR.
- Instead, spend those later round picks on DL and Interior OL
- In the later round pool, RB, LB, TE are better to take in R4 than R7, within the pool of later round players, production is still fairly predictable.
While I believe in these results, I also think than in looking at players who have the chance to be truly elite, exceptions should be made.
Flame away!