Every franchise eventually gets a "greatest draft picks of all time" list, and most of them are the same article: a countdown of the best players the club has had, lightly reshuffled, with the word "steal" doing an enormous amount of unexamined work. It is an enjoyable format. It also answers almost none of the question it appears to be asking, because ranking draft picks is a different problem from ranking players, and a harder one. Buffalo is a good place to see why.
The names nobody argues about
Start with the settled part. The Bills' history includes a cluster of Pro Football Hall of Famers who were Buffalo draft selections and whose careers are inseparable from the franchise: quarterback Jim Kelly, running back Thurman Thomas, wide receiver Andre Reed, defensive end Bruce Smith — the NFL's all-time leader in sacks — and, from an earlier era, offensive lineman Joe DeLamielleure.
No list is going to omit those names. That is the first clue that the ranking is doing less work than it looks like it is doing. Everybody agrees on the population. All the disagreement lives in the ordering, and the ordering depends entirely on a definition of "best pick" that these articles almost never bother to state. State it, and the list changes shape immediately. State a different one and it changes again.
Below are the four reasons the ordering is unstable, in roughly increasing order of how badly they break the exercise.
One: the eras are not comparable, and that is not a quibble
The NFL draft has not been the same institution twice. Three changes matter enough to make cross-era comparison close to meaningless.
The draft has been very different lengths. Seven rounds is a recent arrangement. For long stretches before the mid-1990s the draft ran to a dozen rounds or more, and in its earliest decades to far more than that. A player taken in what was then a middle round might sit, in today's structure, well past the end of the draft entirely — among the undrafted. "Late-round steal" is therefore not a fixed category. It is a description of a position on a scale that has been repeatedly rescaled.
Free agency did not exist for most of it. Unrestricted free agency in anything close to its modern form, and the salary cap alongside it, arrived in the early 1990s. Before that, a club that drafted a great player largely kept him. The draft was not one channel for acquiring talent among several; for practical purposes it was the channel. That inflates the apparent value of every early-era pick, because the counterfactual — sign an equivalent player instead — was not available.
The rookie wage scale changed the economics completely. Before the 2011 collective bargaining agreement, top picks negotiated freely, and the very first selection could command money that dwarfed proven veterans. A high pick was a genuine financial risk. Under the slotted scale it is not: rookie contracts are cheap relative to production across the whole first round. The same selection, at the same number, in the same club, means something structurally different either side of that line.
So when a list places a 1970s pick above a 2010s pick, it is not comparing two decisions. It is comparing two decisions made inside two different games.
Two: draft value is surplus, not talent
This is the point that quietly demolishes most rankings.
The draft's return to a club is not the quality of the player. It is the quality of the player minus what acquiring him cost. A great player taken with the first overall selection is a smaller win than a comparable player taken in the fifth round, because in the first case you spent the most valuable asset in the draft and in the second you spent something close to a rounding error. The player is the same. The pick is not.
Front offices have arrived at this from the other direction. The old trade value charts price picks by expected talent and fall away steeply from the top; later analytical work has generally concluded those charts overvalue the very top of the draft precisely because they ignore cost. Clubs that trade down repeatedly are usually acting on surplus, not on timidity.
You can watch the arithmetic do this yourself. The figures below are illustrative placeholders, not real contracts — put your own estimates in.
Two hypothetical picks. For each, estimate what the player was worth per season on the open market, what he actually cost against the cap per season, and how many seasons you got at that level.
The defaults are deliberately set so the two picks come out identical. A considerably better player who cost a premium selection and a real contract can be worth exactly what a merely good one who cost neither is worth. Any ranking that would put those two in a clear order is ranking something other than the pick.
Three: peak and longevity are different currencies
Suppose you have two selections. One was excellent for four seasons and then, through injury or decline, was finished. The other was very good, never quite excellent, for fourteen. Which was the better pick?
There is no correct answer, only a declared preference, and the preference has consequences.
- Weight peak and you favour players who moved the ceiling of a specific team in a specific window. This is the fan's instinct, because championships are won by peaks and remembered by moments.
- Weight accumulated value and you favour durability, and you will find yourself ranking useful linemen above spectacular skill players, which no list published for readers has ever survived doing.
The two answers also interact with the surplus point above in an awkward way. A long career means the club eventually has to pay market rate — the cheap years end, and every season after that returns much less. A four-year peak that happens entirely inside a rookie contract can be worth more to a club than a fourteen-year career that spent ten of those seasons being paid properly. That conclusion is correct and nobody wants it.
Four: a list of hits tells you nothing about drafting
This is the one that should end the format outright.
A "best picks" list is a list of survivors. It is assembled by looking at which players turned out well and then reasoning backwards to the decisions that produced them, which is exactly the shape of survivorship bias. The denominator is missing. Every one of those selections was made alongside dozens of others by the same people, in the same rooms, using the same methods, and the ones that failed are not in the article.
Two clubs can produce identical top-twenty lists while having drafted completely differently — one hitting on a fifth of its picks, the other on a twentieth. You cannot distinguish those two organisations from their highlights, and the difference between them is the entire subject.
It gets worse. A great player is not proof of a great decision. Clubs have taken players for the wrong reasons and been rescued by a positional switch, a coaching change, or the simple fact that a nineteen-year-old developed in a direction nobody projected. Outcome is a noisy signal of process at the individual level, and a list of individual outcomes is therefore the least informative possible way to evaluate a drafting operation.
What actually distinguishes good drafting organisations
If not the hits, then what? The things that hold up under scrutiny are almost all process rather than personnel.
| Criterion | What it rewards | What it hides |
|---|---|---|
| Top-twenty picks | Star quality and memory | Every miss; the entire denominator |
| Hit rate by round | Consistency of evaluation | Roster context — a good player on a bad team looks worse |
| Surplus value per pick | Cost discipline, trading down | Championship windows, which peaks decide |
| Games started by the class | Durability and immediate usefulness | Quality, and desperate teams starting bad players |
| Second contracts earned | Both club and market agreeing the player is good | Cap-driven decisions that had nothing to do with ability |
Nothing in that table is a ranking of individuals, and the last row is probably the most honest single measure in it: a drafted player whom his own club chose to pay again, at market, is a player two independent judgements have converged on.
Beyond measurement, the behaviours that recur in organisations that draft well are unglamorous. Accumulating picks rather than concentrating them, because the draft is a variance problem and more attempts is the only reliable answer to variance. Treating compensatory selections as a real asset stream rather than a consolation. Being willing to take the highest-graded player available rather than reaching for a positional need that free agency could have solved. Keeping the same evaluation staff long enough for their judgements to be checkable. And having a clear internal view of what the club's own coaching can and cannot develop, which is the part outsiders can never audit.
So what would an honest version of this page look like?
It would declare its definition first: are we ranking players, decisions, or returns on assets. It would publish the misses alongside the hits from the same drafts. It would state whether it weights peak or longevity. And it would probably conclude, having done all that, that the interesting object is not a top twenty at all but a small number of years in which a front office repeatedly got something right — which for Buffalo means the drafts that assembled the team that reached four consecutive Super Bowls, and which is a story about an organisation rather than a countdown.
We are not going to publish the countdown, because we would have to invent the precision to do it, and this site does not do that. What is above is the part we can actually stand behind.
A note on this page's address
This site occupies an expired domain, and this URL previously held a ranked list on the same subject, published years ago by the site's former operator. That article is gone and nothing of it is reproduced here; the page above is new writing. If you followed an old link expecting a countdown, the four problems described above are the reason you are getting an argument instead.