NFL Draft Strategy Condensed Plots

• In the first plot, I am trying to measure how much value a position group brings by a handful of measures

  • Along the x-axis is the absolute value of the Wins Above Replacement gained from improving the worst WAR to the mean WAR at that position. 

    • Essentially, the x-axis is the order of positions starting from right to left that you should improve assuming you were the worst in the NFL at all positions. Wide receivers having a lot of value if they are good, and hurting your team a lot if they are bad means that they provide a lot of value for improving one spot.

  • Along the y-axis is a measure of the positional percentile gained per million expected dollars. This is based on the approach from PFF's Brad Spielberger to approximate NFL contract extensions based on their two year WAR percentile of the top 5 at their position. If you are the best player at your position, you get a 10% pay increase of the top 5 players at your position. If you are about 50% as good as the top 5 players, you should get 50% of their contract. This approximation holds relatively well for actual contracts given out. The y-axis here is measuring a player's two year percentile divided by what their contract would be under these assumptions. Again, if you were the 50th percentile at your position and you would therefore be paid $10 million, your y-axis value would be 0.5/10 = 0.05. We take those values for all players (regardless of their actual contracts or rookie deals) and average across the positions to get the values shown below. 

    • Essentially, this measures how fast the dropoff of the best players at each position is, as well as how expensive the market is. Positions with a high value like HB and OC are ones in which the 10th best player isn't that far off from the top players and the market isn't that expensive either. Positions with low values are the opposite. The dropoff at edge rusher from the top guys to an average player are massive, and even still they get paid a lot because the top end of the market is extremely expensive.

  • Finally, the size of the dot is a measure of how many years on the average rookie contract will that position be at their expected peak efficiency. This is based on work done by PFF's Timo Riske on development curves in the NFL.

    • Essentially, most positions have a rough rookie year and then are good to go for their career. Some positions, especially in the trenches, take longer until they start producing like what the peak of their career is going to be. There will always be players that buck the trend in either direction, and development is not linear, but this gives an expectation of how many years you will likely get of peak production while on a cheap deal. 

• In the second plot, I am trying to measure how difficult it is to find each position group in terms via free agency and the draft

  • Along the x-axis is the average percentage of "star" players (PFF graded above 75) who have changed teams since 2011 by position. 

    • Essentially, how many impactful players actually changed teams since the 2011 CBA to give a sense of where teams can find each position. Important to note is that even the highest positions aren't that high of a percentage. If you want an elite player, regardless of position, you pretty much have to draft or trade for them.

  • Along the y-axis is a weighted average hit rate across the first three rounds of the draft. The weighting gives more importance to the first round hit rate, then the second, and finally the third. For determining if a player was a hit, again I used the cutoff of their NFL PFF grading being above 75 during their career. 

    • Essentially, this gives a sense of how difficult each position is to scout. There are also elements baked into this regarding sample sizes. There aren't that many TE or OC drafted early in the draft, so when one goes that early its often because they are really good. This skews the hit rate very positively, when it doesn't necessarily mean that players of that caliber are available in the early rounds every year. Additionally important to note is that the average hit rate is shockingly low as well. This pairs well with the idea that the draft is mostly a crapshoot and trading down is more advantageous since you probably won't be correct on your evaluation. 

  • Finally, the size of the dot is a measure of the average capital spent on each position group since 2011. This takes into account the draft capital spent, along with the contracts given out to each player. All of this is converted back into WAR to create a unifying metric to evaluate on.

    • Essentially, the most invested in position (other than QB) is ED. They consistently demand early first round selections and garner massive second (and third) contracts. The hit rate is ultimately not that great (though it increases if you only consider the players that go in the top 16 pick). 

College Wide Receiver Metrics:

I have done a decent amount of work on College WR translating to the NFL. I feel like my numbers do a very solid job of predicting NFL success regardless of round taken (though earlier always helps). The major things I look for are as follows.

• Size/Speed Thresholds: Using previous analysis, a WR usually has to be above 5'11" and 195lbs (though that weight has been decreasing in recent years) and run faster than a 4.6 40 yard dash (or equivalent game speed) to be highly successful in the NFL. Players can be very effective and useful for their team at lower height/weight/speed, but when looking for truly special 1,000 yards, #1 dominant guys you pretty much have to be above this threshold or named Tyreek Hill.  

• Separation Ability: The 2019 draft class of WR really messed up my evaluation metrics because a lot of boundary jump ball receivers flopped in the NFL despite having really good college production (Hakeem Butler, Kelvin Harmon, JJ Arcega-Whiteside). One of the common threads with their profile was a high contested target percentage on their downfield throws. If every throw down the field is a contested throw, it means you aren't separating vertically very well. I started converting that into my separator thresholds for college players to see if your seasons are non-separating or not. It is not untenable to be a non-separator, but if multiple seasons or most of your seasons are, that is usually not good. 

• College Efficiency: I created a Value metric based on a player's college efficiency and production using yards per route run, targets per route run, their PFF grades, and other metrics to establish a percentile of college efficiency. There is generally a pretty good trend of being better at Value correlating to being better in the pros, but obviously this is one of the lower importance thresholds unless you are really far down the Value percentiles.

Below is a plot using these metrics and comparing to players that are already in the NFL to show roughly the hit rates by different situations. If a player meets all of my thresholds and is in the solid cluster, they tend to be successful in the NFL (barring off-field or injuries). 

• Players in the Gadget cluster (below the 20th percentile average depth of target in college) very rarely work out in the NFL. 

  • The only true hit is Stefon Diggs from the first season that I have college data for. His previous seasons might have bumped his ADOT and pushed him out of this cluster.

• Players in the Non-Separator cluster (composite metric using ADOT and contested target percentage) tend to not work out, but there are more players who figure it out in the NFL from this cluster. A player landing in this cluster is not a death sentence, but you would prefer they weren't in this cluster. Usually players in this cluster are more athletic and need to refine the rest of their receiving skillset.

  • 5 (Rashee Rice, Tee Higgins, Michael Pittman, George Pickens, Nico Collins) hit and were above the 80th percentile Value and 2 (DK Metcalf, Terry McLaurin) were hits below the 80th percentile out of the 51 non-separators in the dataset. Total (7/51 = 13.7% hit rate)

• Finally, we have the players in the Solid cluster. These players have no obvious issues in my numbers, though I like to install a threshold at the 80th percentile of college efficiency value as it helps avoid a lot of misses. 

  • 13 players hit and were above the 80th percentile Value, 7 more players below the 80th percentile hit out of the 282 total Solid players. Total (20/282 = 7.1%)

  • For only players above the 80th percentile: 13/61 = 21.3%

  • For only players below the 80th percentile: 7/211 = 3.3%

In conclusion, for all of the players who became successful by this threshold in the NFL (28 total), 13 of them (46.4%) came from the Solid cluster and above the 80th percentile (regardless of draft position). 

NFL Comparisons Generator:

Using the above metrics for my wide receivers in both college and the NFL, I added in a projected value for draft prospects based on their most similar comps and how they performed in the NFL.

Let's take Malik Nabers below for example:

Given his college metrics for efficiency, non-separator ability, height, weight, Relative Athletic Score, ADOT, and Slot percentage, here are the top 5 comps in terms of similarity with NFL player's success. To project this player's NFL value, we take the similarity score and mulitiply it by the comparisons NFL actual value which is in the ExpectedWAR column. Once we have that value for the top 5 comps, we take the average of those values multiplied by the average of the similarity scores. This is the final MeanExpectedWAR column. For Malik Nabers, his college profile is very similar to Brandon Aiyuk, Stefon Diggs, and somewhat similar to DJ Moore, Terry McLaurin, and Odell Beckham Jr. Given that those players all had very solid NFL careers, his expectation is correspondingly very high as well. 

Here is a quick list of other draft based analyses that I have completed in the past that I want to touch on here.

• Arm Length for bench press from physics perspective: Defining the amount of work that is done by a bench press in units regardless of arm length. In general, more reps outpaces longer arms unless you are talking about a 3+ inch difference in arm length.

• Combine Analysis for increased WAR: Based on Kevin Cole's work at PFF for defining how much value each combine event has on increasing WAR for a position.

• Hit Rates for Edge Rushers: Edge rushers above a select set of thresholds hit at around 70%, but they usually go within the first 10 picks. 

• WAR per dollar spent on each team: Comparing efficiency versus total dollars spent.

• Draft Capital Investment: Looking at home grown talent rates, as well as what teams invest and hit on.