For forty years the best college football players left the moment they were allowed to. Then somebody started paying them to stay, and about sixty a year quit walking out the door. What you are watching on Saturday is the shape of that decision.
NIL raised the level of play in roughly the top third of FBS. Not by making anyone better at football, but by making it expensive to leave.
Two forces, and they compound. About sixty NFL-caliber players a year who would have turned pro are still on college fields, and the ones who eventually go are half a year older and considerably more used to being hit. That larger pool of talent then drained into a set of programs narrower than any in the last forty-six years. More of it, staying longer, in fewer places.
A junior with a third-round grade used to face a question with one sensible answer. The NFL was the only place that paid. Staying in school meant another year of risking a knee for a scholarship and a meal plan, and the players who could leave, left.
That arithmetic broke in July of 2021. A projected seventh-round pick now signs a rookie deal worth less than what a Power Four program will pay him to come back. And unlike the rookie deal, the school's offer does not require him to beat out four other bodies in August to collect it. For everyone outside roughly the first hundred picks, the exit stopped making sense.
The response was immediate and it was enormous. Early declarations averaged 126 a class from 2019 through 2021. They averaged 70 from 2023 through 2025. The 2026 class produced 63. That is not a trend, it is a repricing: half the door closed in four years, and the players who did not walk through it are still out there on Saturday, a year older than the guy who used to be standing in that spot.
Players getting older is not proof of anything on its own. The pandemic handed an extra year of eligibility to everybody who played in 2020, and that alone would push draft ages up. So the two explanations have to be made to disagree, and they do.
A COVID eligibility year was granted to everyone. If that is the cause, age rises in every round at once. The money story says something far more specific: a first-round pick leaves no matter what anyone offers him, because nothing a collective can pay competes with a guaranteed first-round contract. Only the players further down the board face a real choice. So the increase should appear below round one and nowhere else.
That is exactly what it does.
Rounds two and three rose a quarter of a year. Rounds four through seven rose a fifth. Round one fell, and the share of first-round picks aged 23 or older dropped from 22.9% to 15.6%. The best players are leaving earlier than they used to, exactly as a bidding war for the top of the board would predict.
This is the one place in the whole argument where the word because is earned. COVID eligibility would have lifted all three bands. Conference realignment and a twelve-team playoff do not change what a 21-year-old decides in January. Better coaching would not sort itself by draft round. The money is the only explanation that predicts this particular shape, and this is the shape.
Retention explains why there is more talent in college football. It does not explain why it feels concentrated in about fifteen games a year. For that you have to look at where the talent ended up, and the answer is stranger than the sport's own conventional wisdom.
Take the first 224 picks of two drafts and pin each one to the program it came from. In 1981 and 1982 those 448 names were spread across 147 schools. The tallest single stack on the board was fourteen. Sixty-two programs put up exactly one.
Run the identical exercise on 2025 and 2026. Same 448 names. They now come from 94 schools, the tallest stack is twenty-four, and the long ragged tail of one-pick programs has been cut in half.
The pins did not multiply. They moved. And they moved in a direction that the sport's own folklore gets backwards: this is not the blue bloods pulling away from everyone else. It is the bottom of the sport being emptied into the top.
The two middle bands are the control group, and they are beautifully boring. Programs ranked 26th through 40th have produced between 32.0 and 38.5 picks a year for forty-six consecutive years, and produce 35.5 today. Ranks 41 through 60 have sat between 26.5 and 34.5 the entire time, and sit at 27.0. Nothing happened to the middle of college football. Everything happened at the two ends.
Measured properly, as the effective number of programs actually feeding the league, the last two drafts are the narrowest in at least forty-six years, and not narrowly so. The figure is 47.0. The range across every other two-year window since 1981 is 56.8 to 86.6. It is not at the edge of history. It is outside it.
That last point matters more than anything else in this section, because it forecloses the comfortable reading. You could tell yourself the draft simply got smarter, that scouting improved and the league stopped wasting picks on small-school fliers. The record says otherwise. First-round picks out of non-Power-Four programs returned more career value than first-rounders out of Power Four programs taken in the same range, and washed out less often. The pipeline that closed was producing real players for thirty-five years.
It closed anyway. Not because the market got smarter, but because the market got bought.
A claim about "college football" that does not say which college football is not worth much. So here is the boundary, and it was found rather than chosen.
Rank every program by draft production inside its own window, then ask at which rank the gains turn into losses. Against five separate baselines the answer comes back the same: the crossover sits at rank 39. FBS fields about 135 programs. The line falls, almost exactly, at the top thirty percent.
Above it, cumulative production went from 146.6 picks a year to 178.0. Below it, from 62.6 down to 46.0. The programs ranked 61st and below have watched their share of the draft fall from a quarter of it to 8.5%.
Read that roster twice. Indiana is on it. Texas Tech is on it. Florida State, North Carolina, Wisconsin and Stanford are not. This is not a cartel protecting itself. Against five different baselines, between five and nine of the top fifteen seats changed hands inside a decade. The room got smaller and the door kept swinging.
Texas Tech deserves its asterisk, and gets one: the program produced one or two draft picks a year, every year, from 2018 through 2025, and then nine in the 2026 draft alone. That is a single-year spike, not an established pipeline. It is also the cleanest illustration available of the actual finding, which is that a seat on this board can now be bought inside one recruiting cycle.
Three numbers decide whether any of this survives the decade. They are published annually and none of them require our cooperation.
If early entries climb back toward 100 in 2027 or 2028, retention was a transition artifact and not a repricing. The whole causal claim goes with it.
If this recovers above 15%, consolidation was the portal clearing a one-time backlog of misallocated players, not a structural change.
If the crossover drifts up, first to rank 25 and then to 15, the tier is still contracting, and this page will have understated what happened.
Every honest theory carries its own list of ways it might be wrong, and burying that list would be the fastest way to deserve not being believed.
NIL cannot be separated from the transfer portal. NIL arrived in July 2021, but the portal opened in 2018, one-time transfer in 2021, unrestricted transfer and realignment and a twelve-team playoff in 2024, and House revenue sharing in 2025. That is one natural experiment wearing seven different names. The consolidation section describes what happened; it does not claim to know which rule did it. Only the retention finding in section 01 makes a causal claim, and only because the round-one placebo rules out the alternatives.
The long tail was already thinning. The HBCU pipeline to the NFL declined for decades and was largely finished by around 2015. Alcorn State's last pick came in 1999, Bethune-Cookman's in 2005, Florida A&M's in 2013, and several programs in the early baseline simply stopped playing football altogether. That does not explain the break after 2015, since the effective-schools figure sat between 74 and 76 straight through 2011 while those declines were already underway. But the raw count of pick-producing programs is not a clean talent signal, and should not be read as one.
This counts the program a player was drafted from. It cannot separate a player his school developed from one it acquired in the portal. Arguably the acquisition is the phenomenon. Still, it is a limit worth naming.
Two drafts carry a lot of weight. The current window is 448 picks. Bootstrap intervals say the gap against the 2010 era is real and not sampling noise, but it is a short window and 2027 will tell us more than any argument here can.
And we never graded the football. Everything on this page measures who is on the field and how long they have been there. Nobody here watched the tape. The nearest independent check is Li & Derdenger in Management Science, who find that point spreads tightened after NIL once talent and the portal are controlled for. That is competitive balance, and not quite the same thing as quality.