NFL Draft Value Chart (Updated 2023!)

While watching the first round of the 2022 NFL draft, all I could think about was different ways to construct an NFL draft value chart which was as accurate as possible. Many analysts have constructed draft value charts before, but for one reason or another I think they are each flawed in some crucial way.

I’ll start with the data, and give you my methodology afterwards. I’ll also include a discussion on how my NFL draft value chart compares with other analysts charts.

The Data Jocks’ NFL Draft Pick Value Chart

The full first 7 rounds of picks – assuming there are no compensatory picks – are contained in our NFL draft value chart below.

Another way to view this data is by plotting draft pick value against pick number; doing so yields the following chart.

Before reading onwards, I encourage you to play around with the numbers in the NFL draft value chart shown above. Do they make sense to you? Do they feel wrong?

Jimmy Johnson Draft Chart / Trade Values

The first ever NFL draft value chart was created by Jimmy Johnson in the Cowboys’ front office in the early 90s. This table normalized the first pick to have a value of 3000 and the values decreased exponentially from then onwards. For consistency and ease of comparison, our NFL draft value chart also normalizes the pick values so that the first overall pick has a value of 3000. This way you can compare our values to the Jimmy Johnson draft chart.

Our Chart and the 2022 NFL Draft

We can use our chart to evaluate the winners and losers of trades that happened in the 2022 NFL draft. Below are a few of these trades of picks evaluated using both our chart and the Jimmy Johnson draft chart.

The Washington – New Orleans Trade

Washington Gives No. 11 to New Orleans for Nos. 16, 98, and 120. Here is what each team gave up via our model:

• Washington: 2234 from Pick 11 (2234)
• New Orleans: 2379 from Pick 16 (1944), pick 98 (290), and pick 120 (145)

Our model seems to think this is pretty fair. Because of the opportunity cost of roster spots, usually the team doing a package deal has to give a bit of extra value. New Orleans gave roughly 6% more value to make this work. A great piece of evidence for our model!

The Baltimore – Buffalo Trade

Baltimore gave up the 130th overall pick to move up two spots from 25 to 23. Our model says each team gave up:

• Baltimore: 1785 from Pick 25 (1688) and Pick 130 (97)
• Buffalo: 1744 from Pick 23 (1744)

Again, our model pretty clearly nails this trade. Baltimore gave up about an extra 2-3% of value as part of the opportunity cost price tag to Buffalo.

The Chiefs – Patriots Trade

New England elected to move back in the first round from pick 21 to pick 29. As compensation, the Chiefs offered picks 94 and 121 to even it out. But did it? Each team gave up:

• New England: 1789 from Pick 21 (1789)
• Chiefs: 1941 from Pick 29 (1479), Pick 94 (322), and Pick 121 (140)

Our model is still fairly close to assigning the right values, though it thinks the Chiefs paid slightly more for the deal than New Orleans did. New Orleans paid a 6% penalty in packaging a 3-for-1. The Chiefs paid closer to an 8.5% penalty here.

Was it worth it? That is still yet to be seen. But the main conclusion is that our draft pick value chart works well.

Some Comments on Our Draft Pick Value Chart

If you compare this to Jimmy Johnson’s trade value chart, you’ll see that we generally think that early first round picks are worth less relative to later picks. Purely from a value perspective, the first overall pick with our model is worth roughly equivalent to a package of picks 28 and 29 together. The original draft value chart indicates that if you want to spend back to back picks to secure the first overall pick, you would need to give up numbers 7 and 8 overall. This difference is stark.

Most people’s knee-jerk reaction will be that the old draft value chart ‘feels’ more right. However, I offer two reasons that this doesn’t make the old chart more accurate than our NFL draft value chart.

1. NFL Team’s have historically used charts similar to Jimmy Johnson’s, biasing our expectations

Because Jimmy Johnson’s NFL draft value chart was the first one that was widely used and accepted as accurate, it shouldn’t be surprising that the majority of trades look fair based on the old chart. If the old chart is what was dictating fairness of trades, then of course it will accurately describe the value of picks. Teams likely consulted the old NFL draft value chart when they traded draft picks in the first place to make sure that trades were fair.

If a biased system influences the way data is generated, then that data can’t be used to measure the validity of that system. This idea is regularly discussed in machine learning and data science circles and is well known by the name data leakage.

2. The old NFL draft value chart was designed arbitrarily and without data

The methodology that generated the old trade value chart was quite simple, in fact it was too simple. Jimmy Johnson assumed that the value decayed exponentially. Roughly that means that the percent decrease from one pick to the next was always the same amount. While this isn’t a terrible assumption, we don’t necessarily have any reason to think this is the case. I would argue that the difference between picks 1 and 2 is different from the difference between picks 31 and 32. If you force your model to follow a specific form, you inherently introduce errors.

Moreover, the rate of decrease of the old trade value chart was chosen by feel. It wasn’t designed in a modern way where the curve is fit to observed data. Instead, the curve was chosen in an entirely arbitrary way based on what felt like the correct values.

Because of these two reasons, I suggest that even though our model feels slightly wrong, it actually more accurately describes the true value of draft picks. When you look at how our model matches with the data, you can’t argue with its accuracy.

That is, if it seems like our NFL draft value chart is wrong and doesn’t match with your beliefs, look into the quality of your beliefs. The numbers don’t lie.

How We Built our NFL Trade Value Chart

The NFL draft value chart came about in three steps.

1. We computed the average value that you get from a player drafted at each pick
2. We fit a non-parametric model with quadratic programming that de-noises the data
3. We adjusted the model values relative to replacement level to account for the fact that it is more valuable to have a few very good players than to have many average players.

These steps give us a model which makes as few arbitrary choices as possible, accurately fits the observed data, and takes into account the limited size of rosters and the effect of this limitation on team construction. The result is an unbiased, assumption free NFL draft value chart.

Average Value at Each Pick

The single most important data point that went into our analysis is pro-football-reference’s approximate value statistic. In the simplest possible terms, approximate value is a way to measure how good a player was over the course of their career. It takes into account both a player’s dominance at their peak and their longevity. While not a perfect measure, a player with approximate value 50 was worth about twice as much as a player with approximate value 25.

This statistic formed the basis for our analysis. For each pick number, we computed the average approximate value of all players drafted at that pick. We only used the value that a player accrued for the duration of their tenure with the team that drafted them. For example, to estimate the value of the first overall pick, we first went through all the drafts since 1980 and tabulated all the first overall picks.

Using only the seasons that a player was on the roster of the team that drafted them is the correct choice because players don’t stay with one team their whole career. If they did, draft picks would be more valuable. Other analysts that have done similar work use career approximate value which doesn’t take this into account.

Below is the average approximate value plotted against draft pick. The values on the y-axis are normalized to live on the same scale as our NFL draft value chart where the first pick has value 3000.

Fitting a Model with Quadratic Programming

I said above that any time you build a model with assumptions baked in, you run the risk of making a less accurate model. However, our model builds in one assumption whose accuracy is unimpeachable. Even though the raw data (incorrectly!) says the third pick has been more valuable than both the first and second pick, we know that this is not the case. The third pick isn’t more valuable. It is true, however, that over the last 35 years the average value of players picked third has been higher than those picked first or second. That doesn’t mean this trend will continue in the future.

Instead of fitting a curve to the raw data, we use a slightly different technique. We want to determine pick values that (a) match the raw data as closely as possible and (b) are decreasing as the picks get later. A really great tool to do this is to use quadratic programming. We used quadratic programming in a very similar way when we studied NBA draft pick values. Quadratic programming is a very general technique in which one minimizes a quadratic function subject to linear constraints. For us, the quadratic function is the squared error between the raw data (the scatter plot points) and our model’s assigned value. The linear constraints are that the value of each pick is less than the value of the previous pick.

The result of our quadratic program overlaid over the raw data is shown below.

Notice that our curve interpolates the data quite nicely and is in fact decreasing. Also notice that the rate of decrease of the curve is not constant. If we had assumed a specific form for our model, we wouldn’t have captured as much of the intricacy.

Adjusting Draft Pick Trade Value Relative to Replacement

The last step is perhaps more subtle and is easier to understand with an example. The 200th overall pick is roughly 15 times less valuable than the 1st overall pick from an approximate value perspective. However, I don’t think any manager would trade the 1st overall pick for the first half of the 7th round in the draft. Why? Because even though you’re getting roughly the same total value, the limits on roster size and number of players on the field at a given time means that value concentrated in a few players is better than spreading the value out over many players.

The problems imposed by the inherent finiteness of rosters is a difficult problem to study. The way we attempted to address this issue is by measuring player quality relative to the worst starter in the league. If you subtract the value of the average worst starter in the league from the value of each pick, you get a number which represents ‘expected value above replacement’. This number represents how much better our team would get by substituting them in for our worst player.

We claim this is a fair way of addressing this ‘concentration of value’ concern. We use the softplus function to make sure that no picks have negative value even after subtracting off replacement level.

The chart below shows our adjusted draft pick value chart. The black dots are the average approximate value, the blue curve is the output of our quadratic program, and the red curve is the value adjusted because of roster size limits.

Notice that adjusting values using the replacement value idea doesn’t change the value of early picks but fairly significantly decreases the value of the mid and late-round picks.

Commentary 1: The Pick is not the Player

Especially when talking about early 1sts, you need to remember that the pick is just a means to an end. Picks eventually become players and once spent, the value is entirely dependent on how good the player is. So, for years where there is a clear number one overall pick – Joe Burrow, Trevor Lawrence, Andrew Luck, for example – the first overall pick will be quite a bit more valuable than the average I’ve listed here. If you were trying to trade up for the number one pick in those years, you were paying for the blue chip quarterback prospect and not for the number one overall pick. Our NFL draft value chart is an average.

Commentary 2: Early 1st Round Picks

One could argue that this methodology severely undervalues early first round picks. Not a single NFL team would trade the first pick for picks 18 and 19, much less for picks 28 and 29 which my model deems fair. While we have tried to account for the impact of having significant value in a single player, perhaps we haven’t done that significantly enough. The general team building philosophy is that you need a few stars on either side of the ball and then need to fill in the gaps with average or above average players.

If this philosophy is correct, then early firsts should be more valuable than my model says because the best chance at landing those players is in the early first. I, for one, am not sold on this line of reasoning because the approximate value calculation already places a premium on the best players in the league.

To receive email updates when new analytics come out, use the subscription form below!