The Quarterback Cliff
The draft decides who plays. It does not decide who is good.
212 quarterbacks drafted 2008–2025. Headline analyses run on the 173 mature 2008–2022 picks: 45 Round 1, 34 Day 2, 94 Day 3.
What a quarterback pick actually buys
| Draft day | n | Median career starts | Median attempts | Median starter seasons |
|---|---|---|---|---|
| Round 1 | 45 | 74 | 2,389 | 4 |
| Day 2 | 34 | 17 | 576 | 1 |
| Day 3 | 94 | 1 | 20.5 | 0 |
2008–2022 classes, outcomes through 2025. Median Day 3 QB career: one start. Not one season — one game. Every comparison clears Holm correction.
Hit rate by pick bucket
No magic round exists — a square-root curve fits best and round boundaries add nothing (p = 0.59–0.87). The practical break: picks 1–64 are a different universe from picks 65+. Small buckets (n printed in tooltip) — don't over-read the 6–10 bucket's 100% on n = 8.
Availability is the gate
You cannot finish top-12 on 13 starts — zero counterexamples in 14 tries. Half of the 12–14-start QBs were startable in Superflex; none was ever a QB1. (12–14 row: n = 14 — suggestive, not a law.)
The equality test — once they hold the job, the draft is over
| Metric (16+ career starts) | Round 1 (n=47) | Day 3 (n=10) | p |
|---|---|---|---|
| EPA per dropback | 0.0294 | −0.0181 | 0.356 |
| ANY/A | 6.89 | 6.52 | 0.45 |
| Completion % | 62.8% | 62.7% | 0.877 |
| Fantasy pts per dropback | 0.426 | 0.389 | 0.528 |
| Interception rate | 2.30% | 2.43% | 0.861 |
| Dropbacks per game | 36.6 | 37.7 | 0.78 |
0 of 48 comparisons separate Round 1 from Day 3 at the 16-start bar; 0 of 48 at 400 dropbacks. Once a Day 3 QB holds a starting job, his season is at least as likely to finish top-12 as a first-rounder's.
Rushing is a multiplier, not a substitute
Split at the medians (10.5 rush yds/game, 0.067 EPA/dropback), n = 504 starter seasons. Rushing lifts every efficiency tier — but no amount of rushing rescues a bad passer (OR 0.126, p = 2.2e-14 for legs-over-arm).
Top-12 rate by rushing yards per game
U-shaped, not gated: the 10–15 yds/game band is the worst cell in the study (29.8%) — worse than QBs who don't run at all. Either the offense is built around the arm or the legs; in between is the worst place to be.
Forecast the legs, stay humble about the arm
Year-over-year Spearman correlation, 338 consecutive starter-season pairs. Efficiency has the largest effect on this season's finish and is the least projectable into next season. Rushing is more than twice as stable (0.864 vs 0.413).
The pre-draft board — one signal survives
| Signal family | Signals tested | Survivors (pick controlled) |
|---|---|---|
| Athletic testing | 7 | 0 of 7 |
| College passing production | 10 | 0 of 10 |
| College rushing production | 4 | 4 of 4 |
| Draft age | 1 | 0 of 1 |
| Landing spot at draft | 3 | 0 of 3 |
| Level of competition | 1 | 0 of 1 |
78 tests, six families, each run raw and with log(pick) controlled. College rushing is the only family that survives — and controlling for pick makes it STRONGER. The market has priced everything a scout can measure, and underpriced the one thing that predicts fantasy QB value.
The leash — capital buys forgiveness
| Give-up curve: no top-24 through year… | Round 1 | Day 2 | Day 3 |
|---|---|---|---|
| Year 2 | 28.6% (n=14) | 10% (n=20) | 3.4% (n=87) |
| Year 3 | 20% (n=10) | 5.9% (n=17) | 3.5% (n=85) |
| Year 4 | 11.1% (n=9) | 5.9% (n=17) | 2.6% (n=76) |
After a bottom-tercile season, first-rounders start again 60.3% of the time against 25.0% for Day 3 picks, and the effect survives controlling for performance (p = 0.0004). Play well and everyone keeps their job. Play badly and only the first-rounders do.
The era turn — the modal QB1 flipped
Quadrant composition of top-12 seasons. Passing volume did not increase (31.6 → 31.7 att/game); passing got safer and more efficient, and rushing more than doubled (8.96 → 17.7 yds/game, ρ = 0.951 — the strongest trend in the study).
The 2026 class against the base rates
| Player | Pick | Team | College | Bucket | P(ever starts) | P(ever QB1) |
|---|---|---|---|---|---|---|
| Fernando Mendoza | 1.1 | LV | Indiana | 1-5 | 95.2% | 76.2% |
| Ty Simpson | 1.13 | LAR | Alabama | 11-20 | 88.9% | 44.4% |
| Carson Beck | 3.65 | ARI | Miami (FL) | 65-105 | 36% | 12% |
| Drew Allar | 3.76 | PIT | Penn State | 65-105 | 36% | 12% |
| Cade Klubnik | 4.110 | NYJ | Clemson | 106-150 | 21.1% | 5.3% |
| Cole Payton | 5.178 | PHI | North Dakota State | 151-200 | 5.7% | 2.9% |
| Taylen Green | 6.182 | CLE | Arkansas | 151-200 | 5.7% | 2.9% |
| Athan Kaliakmanis | 7.223 | WAS | Rutgers | 201+ | 5.9% | 2.9% |
| Behren Morton | 7.234 | NE | Texas Tech | 201+ | 5.9% | 2.9% |
| Garrett Nussmeier | 7.249 | KC | LSU | 201+ | 5.9% | 2.9% |
Each rookie shown with the historical hit rates of his pick bucket (2008–2022 classes). Mendoza at pick 1 sits in the single strongest cell in the study; Klubnik's dynasty helium runs ahead of a 5.3% base rate.
The three rules
Rushing peaks youngest (17.9 yds/game under 24) but top-12 rates peak at 26–28 (50.0%). The legs show up first; the efficiency shows up later — buy the young runner while the market prices him cheaply. Corollary: a 30+ QB whose value is built on rushing is a sell.
Stickiness gap, yards minus TDs: +0.403 (CI +0.283 to +0.525). Among top-12 seasons, ~20.8 fantasy points come from rushing yardage and 12.0 from rushing scores. The yardage repeats; the scores do not.
The 10–15 rushing yds/game band posts the worst top-12 rate in the study (29.8%) — worse than QBs who do not run at all (42.1%). Do not pay a mobility premium for a QB who is only a little mobile.
A 12-team Superflex league starts 24 QBs every week. The league produces 26–31 QBs a year who hold a job for ten games, and only 10–18 who make it through a full season. First-round QB capital is the only asset class at the position with a defensible floor, and its dominance is increasing (Round 1 share of starter-seasons rising, ρ = 0.588, p = 0.0103). The QB1-to-QB2 gap has not grown — but rushing now supplies 19–45% of it, versus 2–5% in 2008-2010.