Position Value Theory

It's the offseason. Time for the Combine, scouting reports, free agency, and mock drafts applied to a haze of nearly a thousand "interesting" prospective players. Feeling overwhelmed? Step away from the mind-numbing scrutiny of athletes who are unlikely to ever play for your team, and step into the comfort of my ivory tower.

This is Positional Value Theory. It is a framework of concepts like "scarcity of size" and "frequency of usage". These concepts influence how much a position is valued (as compared to other positions), and how much a better player is valued compared to a worse player (within a position).

So it's not a theory, by which I mean the sort of thing that generates a messy, testable hypothesis. Instead, this theoretical framework will be composed of:

Logic
Truth Value: 100%
Practical Value: 0%

Facts
Truth Value: 100.024% (rounding errors, y'know)
Practical Value: Variable: Either 0% (facts which contradict preconceived notions) or 100% (facts which support them)

Naked Assertions
Truth Value: 0%
Practical Value: Outstanding (you get to challenge them with your own assertions.)

Graphics
Truth Value: pickle
Practical Value: 

The Single-point Failure Model (logic)

Our first theoretical concept involves single-point failure. This can be described by the obscure maxim that "A chain is only as strong as its weakest link." Not all player positions will function like chains, though. To demonstrate the difference let us consider the worst possible player in the NFL.

No, wait. Let's consider a hypothetical player who's even worse. Me.

Prospect Name: Jason Drake
Age: 18 (several decades ago, when my knees still worked)
Height: 5' 7"
Dry Weight: 130 pounds
Strengths: Good dancer. Climbs trees really fast thanks to his long, flexible legs. Versed in calculus. Can juggle.
Weakness: Football.

How is this Drake going to work out? A team's pass blocking follows the single-point failure model. If one offensive lineman fails, it doesn't matter how spectacularly the others succeed. But the opposite is true for their counterparts on the defensive line. They are better described by the popular maxim "A machine gun is just as strong as it's highest-energy bullet." If one pass-rusher succeeds, it doesn't matter how spectacularly the others fail. So plug in our prospect and you get...

A functional complement of pass-rushers:

A very bad pass-blocking line (single-point failure):

Now consider the guys going out for a pass, and the tables are turned. The defensive backs form a chain that is only as strong as its weakest link; and only one receiver needs to get open to make a pass play successful.

A functional complement of receivers:

A very dysfunctional pass defense (single-point failure):

If Drake splits out wide on offense, he can run 10 yards down the sideline and catch a pass if he's completely unguarded. So at the very least, he'll occupy a defender. Put him in at cornerback, however, and he's literally worse than useless, because a clever receiver will use him as a screen to hinder a legitimate defender.

So now we've seen what a General Manager's nightmares look like, but what does this tell us about the value of real players? A lot, actually. It tells us that a football player who forms part of a defensive chain (pass blocker or defensive back) is completely worthless until he reaches a minimum level of competence. And above that competency, his value should level off. An elite left tackle who can shut out 90% of the league's opposing pass rushers is just as valuable as an all-world left tackle (who can shut out 100% of the opposition) against 90% of his opponents. There's no extra value to be had for making it look easy.

On the other hand, a well-below-average wide receiver is more valuable than a terrible wide receiver. Every incremental step in ability translates to a few more routes he can run, a few more defenders he can beat, and a few more passes caught. And there is no upper limit to the value of increasing ability. If you could make a Calvin Johnson 2.0 who's one inch taller and a tenth of a second faster, he'd be more valuable than Calvin Johnson. And you'd be Dr. Soong.

Run-blocking and run defense (naked assertions): Offensive linemen's run-blocking abilities follow both models to some degree. There is a minimal competence required to prevent an opposing defender from bursting into the backfield on a running play, but thanks to the faster timing of run plays this is easier than pass blocking and so does not much inform player value. Run-blocking linemen also have more upside potential; where an attempt to pass block is likely graded as a success or a failure, the ability to not only occupy a defender but push him out of the way has a significant impact on the ability to run the ball, and therefore impacts a player's value.

Although defensive linemen are in the "defender" position on running plays, they do not follow a single-point failure model. It is normally expected that a defensive lineman will be occupied by his offensive counterpart on a running play, hence that is not a "failure" per se. Instead, there is a smoother value curve depending on how much (and how often) such a defender can disrupt the play, either by sideways pursuit, sticking out an arm, or even holding his position despite being unable to make a tackle.

Top Salaries by Position (facts)

Using the latest figures from overthecap.com, I made a salary curve for each of the above position groups. I used the top 32 salaries at left tackle, assuming this would roughly correspond to one per team; and the top 64 salaries at wide receiver, cornerback, and pass-rushing positions (all corresponding to two players per team). Cap allocations for pass rushers definitely followed the "unlimited upside", smooth-curve model. The league's 64th-best-compensated pass rusher (how cool would that be on a business card?) made more money than the 27th-best-compensated left tackle, suggesting that there is a meaningful difference in value between the league's worst starter and the league's best backup (at pass rush). And the curve gets steeper and steeper near the top, with 11 pass rushers making $10M or more (compared to just 5 left tackles), and 5 pass rushers earning more than the highest-paid left tackle.

At the left tackle position, there is a massive jump from #25 ($2.1M per year) to #24 ($4.8M per year), showing a likely point where minimal competence is achieved. Salaries beyond that do not rise so steeply. The highest-paid pass rusher earned twice as much as 73% of the other pass rushers; the highest-paid left tackle earned twice as much as just 44% of the other left tackles.

Now, how about those pass catchers, and their defensive counterparts? Wide receivers had a smooth, up-sloping curve, similar to both the model prediction and the pass rushers.

The data for cornerbacks does not seem to fit the single-point failure model. It looks nothing like the curve for left tackles. But the model can remain fundamentally valid even if it gives very incomplete information about the relative value of corners. With today's passing rules, we can guess that there is one minimum competence level to play corner at all, relying on zone coverage, safety help, and/or a good pass rush to be effective; another minimum competence level where you can play true man-to-man coverage against most receivers most of the time; and yet another level beyond that where we see the so-called "shutdown corners". So instead of one plateau on the salary curve, we see several.

It can even be shown mathematically that the wide receiver curve is more smooth [footnote 1] than that of cornerbacks. Ergo, if these salaries reflect player value, then every small increment in a wide receiver's ability corresponds to a small increment in his value. But cornerback values are flat over small increments, with big jumps occurring at specific points corresponding to different types of minimum competency.

Development and Evaluation Bottlenecks (logic)

How many times have you heard this refrain in player evaluation?: "Yeah, he kicked a lot of 50-yard field goals in college, but he was playing against weak competition. He's not gonna perform like that in the NFL."

The sad plight of kickers is not rooted in their on-field value or limited athleticism, but the simple fact that they are easy to find. For the price of a few footballs and a regulation goal post at its training facility-- neither of which counts against the salary cap-- an NFL team could bring in 10 potential placekickers and give them literally an entire season's worth of full-scale testing in less than a week.

The best kickers really do provide an advantage, and are worth $3M a year or a mid-round draft pick. But an abundance of easily identified talent causes supply to overwhelm demand about halfway down the list. Exactly 16 NFL kickers have more than $50K per year of guaranteed money.

Quarterbacks are at the opposite extreme, and subject to the most severe bottleneck.

Seattle Seahawks defensive end Greg Scruggs never set out to be a football player. But nobody overlooks an athletic 300-pound man, as Scruggs explains: "The high school football coach tried to get me every year, but I wouldn’t do it. I finally wanted to get a scholarship to college, so I went out my senior year."

Now imagine this conversation:

Player: "Coach! Coach! There's a guy in my sociology class who's 6'2", above-average intelligence, and runs a 4.7 40-yard dash!"
Coach: "What a find! We gotta get this guy in here and make him our starting quarterback!"

Yeah, right.

The baseline physical and mental attributes required to be an NFL quarterback are fairly common, certainly in comparison to the those required for other positions. But there is simply no way to learn quarterback skills without being a quarterback in a full-scale game. And even with experience in college or an alternate league (CFL, Arena Football), it's very difficult to predict how a quarterback will perform at the NFL level.

Bottlenecking is most extreme in driving up the value of experienced/known quarterbacks. Its effect varies at other positions:

Offensive Line

The biggest guys on the field depend on the smallest movements: a hand moving an inch to the outside, a foot turned three degrees, or an imperceptible shift in balance. You cannot see these things on film, you cannot measure them at a Combine, and you cannot practice them without full pads and aggressive defenders. Pass-blocking skills, in particular, are very reactive and thus can only be developed at full game speed.

To top it off, it's generally a bad idea to rotate offensive linemen in a competitive game, and a weak link at the position (who's trying to gain experience) can wreck your team's performance.

I read somewhere that successful offensive linemen in the NFL almost always come from major programs. Perhaps a kind reader will supply a reference (I couldn't find it). However, this makes perfect sense, as the necessary skill set can only be developed and seen against better competition. Certainly, the percentage of players drafted from major programs is highest among linemen:

Drafted players who played in the NFL,
percentage from BCS conference schools,
2007 - 2014 drafts

Defensive Line 220/344 = 64.0%
Offensive Line 198/315 = 62.9%
Linebacker 148/239 = 61.9%
Running Back 101/168 = 60.1%
Wide Receiver 145/249 = 58.2%
Defensive Back 216/400 = 54.0%

Wide Receiver

My initial thought is that wide receivers should be easy to find and develop. Most teams rotate second-string wide receivers into the lineup even in regular-season games. More importantly, nearly everything a wide receiver has to do well can be done in a pads-free, less-violent practice environment (the reality T.V. show Fourth and Long used wide receivers and defensive backs).

On the other hand, a wide receiver in a real game has very limited opportunity to test his most important ability, namely, catching the ball. Consider that a rotational defensive lineman who plays 30 snaps will engage in hand-fighting, shoving, and pushing on all 30 snaps, providing a meaty course of development and coaches' film, whether he's in position to make a tackle or not. A receiver playing the same 30 snaps might have just 2 or 3 opportunities to catch a football.

What's more, there are limited reps (both in practice and real games) with the starting quarterback. There's little value in getting open if you aren't seen, or running a perfect route to a misplaced ball; and over-adapting to bad quarterback play could actually be counter-productive.

My conclusion is that wide receivers do not face a difficult evaluation bottleneck, but a more constricted developmental bottleneck.

Defensive Front

This is something of a mixed bag. Like offensive linemen, they cannot practice and display their abilities without violent, game-style contact. But there are a couple of advantages.

First, you can tell a lot about a pass-rusher based on speed and power. While there is no substitute for actual competition, Combine-style measurements are more useful than they are for a pass-blocker.

Second, you can rotate the defensive front seven a lot more than the offensive front. Fresh legs can compensate for a lack of skill, one player's failure to reach the quarterback isn't a guaranteed disaster, and thus your second-string players gain experience and provide film for coaches.

Playing run defense, especially at the line, goes the other way. A defensive lineman reacting to a run-block is in the same position as an offensive lineman reacting to a pass rush, requiring highly-developed split-second shifts of balance and minute turns of hand and foot to gain advantage. Such skill are undoubtedly a factor in allowing guys like Vince Wilfork, Kevin Williams, John Abraham, and Julius Peppers to be effective into their mid-30's.

All the Rest

Running backs are best tested with full-contact play, but there are plenty of opportunities for non-starters to play rotationally.

Defensive backs certainly have to develop skills, but they can do a great deal without full contact. Because a cornerback's first task is to play the receiver, they depend less on having a first-string quarterback to test them in practice.

Punters and long-snappers join kickers at the bottom of the list; they can learn and display critical skills with a minimal surrounding cast.

Physical Rarity (facts and assertions)

Physical attributes such as speed, strength, and size tend to be distributed on a bell curve, with a vast majority of individuals falling close to average. Nature produces very few very special bodies.

Distribution of strength is poorly documented, and weight data would not be informative about natural body types. However, there is good information on height distribution. Among men, the average height is 5'10" with a standard deviation of 4 inches. From this, it's relatively easy to calculate the percentage of individuals who reach a certain minimum height:

22.7% of males will be 6'1" or taller
10.6% will be 6'3" or taller
4.0% will be 6'5" or taller

Going solely by height, there should be plenty of athletes who can play wide receiver. 23 out of 43 receivers who caught 60+ passes in 2014 were 6'1" or under. By contrast, 85% of offensive tackles who started 9+ games in 2014 were at least 6'5". A difference of just one standard deviation (4 inches) makes the requisite attribute five times as rare. We can guess that similar distribution and rarity applies to other peak physical requirements. Breaking it down by position...

Linemen

Offensive linemen must be massive, and tackles in particular must also be very long-limbed in order to defend the edge. And there is simply no substitute for this size, making for a limited cross-section of the population that can potentially play these positions.

Defensive linemen have to contest their counterparts in terms of size, and they need to be fast enough to close on a quarterback or running back. This double extreme creates the most restricted talent pool of any position. Such is often cited as a key motivation in developing the 3-4 defensive scheme, which compensates for the shortage of athletes by making use of undersized "tweeners".

Middle backfield

Running backs require exceptional strength, particularly in the lower body. However, the pool of athletes who can potentially provide this is large, because they can be average height and require only good (not elite) speed.

The defensive middle-- linebackers and safeties-- must be excellent athletes at the NFL level. But their ability to make plays depends on a combination of speed, acceleration, and mental reaction time, so if they fall short of elite in one attribute they can compensate with the others.

Corners and Receivers

Cornerbacks need speed, acceleration, and vertical agility. And because they line up one-on-one against a variety of receivers, corners cannot fall short in any one attribute. A corner who doesn't have outstanding straight-line speed will get beat deep, a corner with sub-par acceleration will lose his man on cut routes, and a corner who cannot play vertically will be victimized by jump balls.

Receivers absolutely must have the coordination and agility to catch a football. They should be as athletic as possible, but unlike corners they do not require peak ability in every area. The offense dictates the plays and the route types, so a wide receiver can get away with being quick (good acceleration) or fast (elite top speed) or outstanding at competing for the ball against tight coverage.

Versatility and Exploitation (logic, et. al.)

This is a complicated blend of related concepts. Let's start with the familiar idea of stretching the field, both vertically and horizontally. This is how I picture the "relative influence" of offensive manpower from a Weak I-Form versus 4-3 personnel set:

The green bands on the outside (best zones for the offense) represent areas where little or no help is available to the cornerback. This fades to yellow in the deep part of the field because, regardless of potential safety help, exploitation of such zones is dependent on pass protection.

The bubble in front of the line of scrimmage is favorable to the offense despite an apparent parity in nearby manpower because this is a measure of relative influence. Offensive linemen typically cannot run very far, and are not allowed downfield on a passing play prior to the ball being thrown; so this bubble shows where 5/11 of the offensive players have nearly all of their effectiveness beyond the line of scrimmage.

This concept explains why lightweight, high-speed "scat backs" are less than ideal. If your running back cannot be effective between the tackles, the opposition defense can abandon that green bubble. The front seven will spread out wide to contain such a back, and in the process be more effective both rushing the quarterback and defending the pass.

Speedy receivers who stretch the field vertically are useful, but they can be neutralized by conservative safety play or a quick pass rush. More coveted are the big-bodied wide receivers who can out-jump or out-muscle a single defender 10-15 yards down the field before a pass rush or an extra defensive back is able to disrupt the play.

A slot receiver, in theory, is much less valuable. Routes to the inside take him where the defense is already strongest, and do not diminish their ability to be in position for stopping the run (the same is true for fullbacks and tight ends on middle routes). Routes to the outside bring an extra defensive back to that side of the field, which creates an opportunity for the defense to play zone coverage and/or have a single safety on one side helping against two (instead of just one) receivers-- in short, the extra receiver interferes with the manpower advantage that the outside receiver should ideally have.

In practice, a three-receiver set is far more popular than using a fullback in today's NFL. And the value of a particular slot receiver's talent to his team will probably depend a lot on what you have at quarterback, running back, and the outside receivers.

An important takeaway from this is that there are five players-- the running back, the receivers split wide, and the outside corners-- who really only need to be good at one thing. A top-notch runner who pulls in the defense does more to help the passing game than a mediocre runner who's good at catching the ball. An outstanding cornerback who can play effective man coverage, thus allowing the strong safety to form an eight-man front, does more to help the run defense than a mediocre corner who is outstanding at tackling a running back 20 yards down the field.

Which brings us to...

Exploitation of Skills

The prestige of the running back position has dropped off dramatically in recent decades. It is now considered a very fungible position, easily filled with mid- to low-round draft picks. The rising effectiveness of passing is cited as the main cause, but I think it has as much to do with better strategic analysis and better tackling.

Consider that a featured running back might carry the ball 15-20 times per game, but his personal skill level won't change the outcome on every one of those plays. If the blockers open a big hole, a replacement-level player could go through it just as easily; if the play is stuffed by multiple defenders, a replacement-level player and a Pro Bowler will both be stuffed for no gain. Even Marshawn Lynch-- as good as it gets-- can only be expected to break a tackle on one carry in six.

The numbers bear this out. The 32nd-best running back in 2014 averaged 4.05 yard per carry. The top three backs averaged a collective 5.22 yards/carry, which is a 29% increase in production.

The 32nd-best quarterback in 2014 had a 4.82 ANY/A (adjusted net yards/attempt). The top three QB's averaged 8.19 ANY/A, a whopping 70% more production.

The disparity is so large not because pass plays go for more yards (note that I used a percentage); it's because quarterbacks have to use a much bigger skill set every time they pass the ball: reading the defense pre-snap, deciding on a target, avoiding the pass rush, timing the pass, and throwing the ball accurately.

Wide receivers fall somewhere in between. Although traditional statistics don't give a ready answer, Football Outsiders calculates "yards above replacement" for running backs and wide receivers. The aforementioned Marshawn Lynch had, by their measure, 362 yards above replacement on 280 runs, or 1.29 yards/run. That's definitely worth something, but nowhere near the gap achieved by top receivers. Randall Cobb had 478 DYAR on 127 pass targets (+3.76 yds/target) and Antonio Brown had 553 DYAR on 181 targets (+3.06 yds/target). Both played with very good quarterbacks, but even the Giants' Odell Beckham scored 394 DYAR on 130 targets (+3.03) and Washington's DeSean Jackson got 310 DYAR on 94 targets (+3.30).

Simply put, the chance to break a tackle on a small percentage of touches doesn't compare with the chance to catch or drop a pass on a very large percentage of targets.

Corollary to this is the frequency with which a player is on the field. Offensive linemen, for example, are neither rotational nor situational, which means their primary skill of blocking is a potential factor on every offensive play. Which brings us to...

The variability of versatility

If a player can help on special teams, his value his higher. This requires no further analysis because the special teams contributions happen across an independent set of plays.

But what about starting-caliber players who can do more than one thing from the same position? Seahawk fans may fondly remember the great fullback John L. Williams, who provided pass-blocking for Dave Krieg, run-blocking for Curt Warner, and along the way accumulated more combined rushing and receiving yards than 17 Hall-of-Famers.

But this diversified arsenal of abilities doesn't make him Superman; it makes him Ultra Boy.

Able to use just one super power at a time, a fullback or a tight end has value by creating defensive uncertainty. If their role were known before the snap, the team would be better served with a specialist (running back, offensive lineman, or wide receiver).

But if we understand this receiver vs blocker conundrum as a Normal-form Game, we should know that a player's value doesn't depend on his best or worst skill set, but a sum of the two. Being a great receiver and an average blocker is just as effective as being above-average at both. You help the run by forcing the defense to play the pass, and if they don't make the right strategic reaction, you "play the grid" to favor the pass.

So tight ends are valuable to have and use, but the dual-threat paradigm changes the talent pool. Imagine, if you will, that wide receiver ability is determined by a single 12-sided die. Going from a '7' to a '6' to a '5', the number of quality players quickly diminishes. But tight end abilities would be determined by 2 six-sided dice (one for receiving and one for blocking), giving a limited number of elite players but creating a glut in the 5-7 range because of the different possible combinations. So tight ends in the mid- to low-skill ranges are not highly valued simply because they are more abundant.

(Note that "elite" is relative to the position. No tight end can actually block as well as an offensive tackle and catch passes as well as a starting NFL receiver. Hence their value peaks somewhere below the best wide receivers and offensive linemen.)

Within their position, fullbacks should follow much the same model as tight ends. But in the modern game, it's better to have such a player at the line of scrimmage than in the backfield because of easier passing rules, faster game speed, improved quarterback play, and modern scheming.

Versatility on Defense

The Ultra Boy on the other side of the ball is the mid-sized linebacker. Versatility on defense is, if anything, valued even less. Simply put, the opposition will attack your weakest ability, not your strongest.

Consider the following list of 4-3 outside linebackers (along with what their teams paid to get them) who recorded 60+ tackles and 4+ passes defensed and 3+ sacks and at least one interception in any single season from 2012-2014:

Lavonte David (drafted 2-58)
Mychal Kendricks (drafted 2-45)
Thomas Davis ($4.8M per year)
JoLonn Dunbar (UDFA)
Akeem Ayers (drafted 2-39)
Chase Blackburn ($1M per year)
Sean Weatherspoon (drafted 1-19)

Those are some respectably high draft picks, but not a lot of salary. By contrast, a 3-4 outside linebacker plays more like a defensive end. He is less versatile, but able to perform essentially the same task on every down, whether it's a running play or a pass play. Among linebackers with 7+ sacks and fewer than 30 tackles and fewer than 3 passes defensed, in any one season from 2012-2014:

Clay Matthews ($13.2M per year)
Aldon Smith (drafted 1-7)
Elvis Dumervil ($5.2M per year)

And it's not just about the pass rush being especially lucrative. Middle linebackers without a lot of pass defense responsibility are also valued for a single skill which they can employ on every down. Among inside linebackers with 85+ tackles and at most 2 sacks and at most 4 passes defensed, in any one season from 2012-2014:

James Laurinitas ($8.3M per year)
Curtis Lofton ($5.5M per year)
Demeco Ryans ($7.8M per year)
Derrick Johnson ($5.5M per year)
Lawrence Timmons ($9.6M per year)
Bobby Wagner (drafted 2-47)

Draft History (facts) and Perspective

In addition to salary figures-- which carry uncertainty because of rookie contracts and cap inflation-- a history of recent drafts can provide some interesting corroboration about the value teams apply to different positions.

The following charts show the number of players drafted at each position in each round. To make the chart easier to read, the count for each position/round is scaled to an estimated roster size.

these numbers aren't important

For example, I estimated 4 running backs for a typical 53-man roster, and there were 14 taken in the 2nd round over the sample, so the chart shows 14/4 = 3.5 RB's for the 2nd round.

Defensive Draftees, 2011-2014, rounds 1 to 6

Defensive linemen are targeted early, with the high-upside, pass-rushing defensive ends dominating the first round. Defensive tackles peak in the 3rd round, then fall off dramatically once the physically-qualified candidates are all taken. A few more physically viable candidates remain at end, and most of these are snatched up in the 5th round.

The curve for defensive backs is much harder to read (the data from PFR does not distinguish corners and safeties). I believe the dearth of activity in early rounds, despite the high value at the position, indicates a difficulty in identifying pro-quality corners based on college performance. In the 4th and 5th rounds, teams are willing to spend a lot of draft picks, presumably taking chances on guys who meet the minimum speed requirements.

Offensive Draftees, 2011-2014, rounds 1 to 6

Receivers, tight ends, and running backs have eerily similar curves, all peaking in the even-numbered rounds. This looks to be a draft-day phenomenon corresponding to a general run on offensive linemen and defensive tackles in the 3rd, and defensive ends and defensive backs in the 5th. The standout among these skill positions is wide receiver-- predicted to have the highest upside for elite skill, it is the most coveted position in the first round.

Aside from the first-round run on pass rushers, offensive linemen have a curve very similar to that of defensive linemen. Viable candidates at the position are more easily identified, based on physical attributes and college experience; but there is a very limited pool available. Here are some numbers which bear that out, looking at players who've been in the league 5-9 years at a few select positions:

Note that the drop-off in wide receiver success in later rounds is not due to the same lack of depth and sleeper candidates as occurs on the offensive line, but to a much higher volume of drafting compared to the actual needs at the position.

And this should help Seahawk fans understand why our team has invested so much draft capital in the offensive line (and, for that matter, free agent money on the defensive line) instead of seeing Schneider's late-round magic. Big and strong guys just don't slip past other teams' notice.

A Completely Arbitrary Summary (naked assertions)

The table below gives an estimate of how much each position's value is affected by each attribute. The attributes are not assumed to be of equal importance. Higher numbers indicate more scarcity and/or more value.

[a] Higher number for outside receivers who have sideline playmaking ability

[b] Higher number for outside corners, lower number for nickel corners

[c] Higher number for running backs who have interior running ability

Physical Rarity refers to natural speed and size, as well as more specialized factors such as arm strength, leg strength, and the eponymous "fluid hips."
Bottlenecks are self-explanatory.
Exploitation Frequency estimates the frequency across different plays where the position's skills are useful. It is lower for mixed-skill positions that use different skills on different plays (such as blocking/receiving tight ends and coverage/tackling linebackers). Positions which attack/defend the isolation zones (outside receivers, quarterbacks, corners, between-the-tackle running backs) are rated slightly higher because of the strategic threat, even if they're involved in a lower frequency of plays.
Upside Skill Value is lower for positions where exceptional skills do not necessarily add any value even when some skills are used. This includes offensive linemen on the single-point-failure model, running backs who sometimes have no lane available, and linebackers/safeties who occasionally make easy tackles.

[1] Receiver and cornerback salary curves: The average interval between consecutive salaries does not tell us anything about the shape of the curve overall, because it can be calculated based on nothing more than the highest and lowest salaries. E.g.,

Mean interval:
WR 1 to 64 $226,992
CB 1 to 64 $226,379

So the total range of salaries is almost identical. However, if we look at the square root of each interval (between consecutive salaries), the smoother curve will have a higher average.

Mean square root of interval, all 64 player salaries:
WR = 357
CB = 335

Mean square root of interval, 1 to 32 (lowest-paid half):
WR = 233
CB = 184

Mean square root of interval from $2M salary to $4M salary:
WR = 247
CB = 213