The 2023 NFL Draft is finally here. On Thursday night, sequences of picks and trades that have been played out in our heads for months will finally become a reality. Thomas Dimitroff and I will be hosting Sumer’s first live draft stream, which you can find on YouTube, Twitch, and Twitter.
Before the events of Thursday night capture the hearts and minds of the football universe, I would like to go through some analytical proverbs of drafting that have crystalized over the past few decades. Some of these have been known for a long time, thanks in part to the seminal work of Cade Massey and Richard Thaler, while others have been developed more recently by some of the best minds in the space.
This list is meant to be instructive not exhaustive. For feedback, @SumerSports on Twitter is a good place to start.
It is really difficult to outdraft the competition pick for pick.
This is hard to admit sometimes given all the work that goes into player evaluation, but the prevailing research suggests that, after adjusting for the expectation of the pick, the broader prediction market is efficient when it comes to ordering players. Arif Hasan of Pro Football Network and Benjamin Robinson of Grinding the Mocks have done great work aggregating the opinions of mock drafters for years now, and the evidence suggests the wisdom of crowds applies to selecting players at the NFL level.
My former colleague at PFF, Timo Riske, demonstrated the efficiency of the consensus approach here, showing that the NFL and the consensus big board coincide roughly 80 percent of the time. Where do the discrepancies arise? Well, one source is captured in the following proverb:
There is really no such thing as a steal, but there is empirical support for the existence of reaches.
One of the differences between NFL teams and the public is access to information like medical and psychological evaluations. While there has been a lot of discussion regarding the S2 Cognition Test in recent weeks, players generally do not skyrocket up draft boards because of medical or psych evals. Instead, they are more likely to fall down a team’s board because of negative information stemming from one of these evaluations. While this information might eventually get to the public, it often stays concealed among the 32 NFL clubs providing a distinction between the actual marketplace and the public one.
To this point, Riske found that between the two extremes, reaches (where a team takes a player higher than their spot on the consensus board) and steals (where a team takes a player that has fallen lower than their spot on the consensus board), there is really only empirical evidence for the former. Teams that take a player higher than expected to are, in a sense, “a market of one”, bestowed with all of the statistical properties of a draft pundit being quoted by Freezing Cold Takes. On average, players that are taken ahead of their consensus position underperform.
Players who fall, on the other hand, have been assessed by some of (if not all) of the NFL marketplace, a marketplace that has access to the tools like medical and psych evals that are not widely available to the public, leading to a market with greater efficiency than the market for a reach. While there are certainly exceptions, players that fall play to the average of their draft position.
The Jimmy Johnson chart is outdated, but it also sets the market. For that reason, trading back almost always generates more value than trading up.
It is a commonly known fact among NFL analysts that the Jimmy Johnson trade chart is not “right” in that it overvalues higher picks and slopes too steeply. This is to be expected since when Mike McCoy created the chart for Johnson in the late-80s/early-90s, draft picks were paid more in accord with the market value of top players at their position, and the chart was derived using the cost of each pick. Still, the Jimmy Johnson chart is the starting point for most trade discussions.
In Massey and Thaler’s paper, they coined the phrase “Loser’s Curse” to describe the empirical fact that the top picks were not generating value relative to their cost (i.e., surplus value) as much as picks later in the first round or even in the second round.
While the new CBA lowered the cost of draft picks and elicited more voracious trades as a result because teams did not need to trade picks and also back up the truck to sign the player, the surplus value of the very top picks are still not that of picks later in the first round as Ben Baldwin showed here.
This makes the price of trading up, when using the prices bestowed by the Jimmy Johnson chart (and most of the other charts), much higher than the value generated. Of course, different teams are in different stages of their competitive cycle, which leads to the next proverb:
Some teams are trading for value while others are trading for surplus value, and that might be okay.
The Jimmy Johnson curve is based on the value of the pick rather than the surplus value. Due to the before-stated efficiency of the NFL draft market, this curve is decreasing and concave up. The surplus value curve exhibits no such shape. Value and surplus value also apply to players relative to their contract, as I discussed in this article on DeAndre Hopkins last week.
As analytics have improved across the league, more teams can ponder if they should trade up for value or surplus value. Many of the analytically inclined believe the latter, that trades should be made to optimize for surplus value and not value. In the long term, this is very likely the right move to make. If a team is buying asset after asset that, on average, generates more than it costs, the positive value is going to add up.
The issue is that many teams do not have the long-term horizon necessary to let the math do its work. Some team builders need to generate a lot of on-field value (see the Jets with their recent trade for Aaron Rodgers), and future considerations are to be dealt with once they emerge from the basin of hot-seated executives. If the acquisition fails, so be it, that is a future executive’s problem. If the bet hits, there is (in theory) time and goodwill to smooth out costs.
In the extreme, see the current Rams and Buccaneers dealing with those future costs, which might be difficult even if it was all worth it because of the championship earned. However, there are instruments, like restructures and void years, that one can use to ease the cost burden post success.
Thus, different assets might have different utility depending on the statuses of the teams, which leads to the next proverb:
If you are going to trade up, make it for a quarterback, but if it is for another premium position, make it a pick swap.
The NFL’s Director of Football Data and Analytics wrote a great piece a few years ago blending different approaches to draft curves to account for landing “stars”, which are more likely to be obtained near the top of the draft. In their mock draft simulator, PFF makes a special consideration for trading up for a quarterback since they have outsized effects on team-level outcomes relative to other positions. The fact that some teams (although fewer than believed) do not need a quarterback while others do can lead to mutually beneficial situations for trade partners.
Recent work by Kevin Cole of Unexpected Points showed that the curves for players that play premium positions – positions like quarterback, offensive tackle, edge, wide receiver, cornerback, and, increasingly, interior defensive line – slope steeper than those of non-premium positions like running back, safety, tight end, and linebacker.
If a team needs one of those premium positions, a trade up might be more of a reasonable play than it would be on average since they are surrendering picks that have less of a chance to turn into impact players at those spots. A team can play multiple premium position players at once (outside of quarterback), and therefore every team needs premium position players all the time. This creates more situations where a trade is more mutually beneficial for both parties than when considering the average situation.
Consider the final point of the proverb; if you are going to move up in the draft, try to do so using what is called a “pick swap”. For example, when Detroit traded picks 32, 34, and 66 to move up to pick 12 in the 2022 NFL Draft, they got the Vikings to give them back pick 46th pick, which is a relatively valuable asset that helped alleviate the problem associated with having the correct “change” in a deal. Usually, the team trading up, often the team initiating discussions, must “round up” due to not having the exact draft picks implied by whatever chart they are using. Swaps help mitigate the rounding error.
The draft is the only cost-effective way to get players at premium positions, so try not to use high draft capital on non-premium positions.
While the definition of “premium position” had for a long time been ambiguous, Jason Fitzgerald of OverTheCap did a great job illuminating the concept last year. In short, a position is a premium position if it is practically impossible to access top-end talent at the position without the use of high-end draft capital. Put differentially, a position is not premium if you can access high-end players at the position through free agency.
For example, among the highest-paid quarterbacks in the NFL, only the lowest-paid of the top 10, Derek Carr, was available as a free agent:
On the other hand, among linebackers, four of the top ten players at the position were accessible in free agency (Edmunds, Mosely, Oluokun, and Okereke), and Deion Jones was available via a low-end pick swap.
Thus, when it comes to building a roster, it is very difficult to find success if high-end draft capital is used on non-premium positions relative to premium ones. Every high draft pick that is used on a guard cannot be used on a wide receiver or edge player, while every high draft pick that is not used on a non-premium position can generally be supplemented by free agent acquisitions.
Additionally, the surplus value of a premium position player – if they hit – is far greater than that of a non-premium position, which provides the surplus necessary to acquire high-end players at non-premium spots. Roughly, hitting on a wide receiver at the top end of the draft can be worth $20 million in surplus value (assuming ~$30 million value), while the same hit on a tight end can be worth roughly half of that (assuming ~$20 million in value). When the Saquon Barkleys and Quenton Nelsons of the world were drafted, they were immediately among the top-paid players at their positions. Joe Burrow was paid $10 million less APY as a rookie than Panthers starter Teddy Bridgewater during his first season in the Queen City.
Furthermore, high draft capital spent on non-premium positions often begets having to use even more future draft resources to acquire premium position players. If a team needs to acquire a great quarterback, the price is generally multiple high picks and a big contract (see Russell Wilson, Matthew Stafford, and Deshaun Watson). This also holds true for other premium positions (e.g., A.J. Brown, Tyreek Hill, Frank Clark, and Jalen Ramsey).
When a team hits on a player at a premium position, the surplus is enormous when they are on a rookie deal and roughly zero when they sign their extension with the team. When a team must buy one in the (trade) marketplace, that is almost never true.
Draft for needs that will exist a year from now instead of needs that are immediate.
Finally, the last draft proverb to discuss is the idea of needs. Needs factor into all draft availability models and are a cornerstone for many of the prices that comprise the betting markets for the draft.
While needs are very much real, drafting for need is generally a fool’s errand. Restricting one’s market to a smaller set of positions increases the chances of a reach, and reaches do empirically do worse than other selections. Additionally, using prediction markets as our guide, rookie impact is – on average – minimal at best.
The NFL betting markets don’t really change much from before the draft to after, unless a team unexpectedly takes a QB they plan to start or trades/trades for a veteran star (PHI). Small changes for things like trading up or back 20 picks in the first round (DET/MIN).
— Eric Eager 📊🏈 (@ericeager_) May 9, 2022
Example. #Chiefs had what was universally praised as a great draft and their SB odds are the same – haven’t changed since they traded Tyreek Hill:
Packers didn’t draft a WR in round one (!) and unlike George Costanza nothing moved https://t.co/YQv6fQbRBW pic.twitter.com/5n2Mz5hWa8
— Eric Eager 📊🏈 (@ericeager_) May 9, 2022
In other words, while it is true that some rookies will have an impact the year they are drafted, it is very difficult to predict which set of rookies will comprise that group. A team using the draft to fill needs for that year are essentially saying, “I know which rookies will have an impact”. History would doubt such a claim.
Given the price of retaining premium position players, even those that are league-average starters, drafting them high a year earlier than you need them is prudent. For example, when one of your tackles is coming up for free agency in year n, it is likely a good investment to draft one in year n – 1, even if the politics (especially at quarterback) can get messy.
The NFL Draft is a revealing event that allows us to look into the soul of teams and their team builders. While none of these proverbs are a silver bullet and some of these edges have partially dried up as teams have gotten more sophisticated in their approaches, in a short amount of time they will be the table stakes upon which we generate new insights for the NFL’s premier offseason event.