Three Definitions of a Bust

Why “bust” needs three definitions, not one

Every fan has a definition of a draft bust. The problem is that everyone's definition is slightly different, and the differences matter. A player who never made a Pro Bowl might still have been a perfectly serviceable starter for a decade. A player with an outstanding career somewhere else might have done nothing for the team that drafted him. A player who was injured after two great years might be classified differently depending on what you care about.

Issue #25 sidestepped this argument by running three definitions in parallel rather than picking one. Each definition captures something real:

• Definition A — "Never made a Pro Bowl." The harshest, simplest test. Pro Bowl is a peer-judged outcome metric — either a player was elite in a season or he wasn't.
• Definition B — "Career value below the slot's expected value." A pick at #5 should produce more career output than a pick at #25; comparing each player against what his slot historically delivers is a fair test.
• Definition C — "Fewer than four seasons as the drafting team's primary starter." Asks the team-specific question: did this pick justify itself to the franchise that spent it?

The "triple-bust" rate (28% over fifty years) is the intersection of all three definitions, not the union. That's deliberate. The intersection is the strict test: it only counts players whose careers are unambiguously below replacement-level by every available lens. It produces a smaller, more defensible number than any single definition would.

From three definitions to a misery index

Each definition is a binary classifier. For a given player, each definition returns a 0 or a 1: did the player meet this bust criterion or not?

The "Expected AV(slot)" function is a slot-by-slot lookup based on Chase Stuart's published expected-AV-per-pick research. Picks earlier in the first round historically yield higher career AV; the curve is monotonic but not linear. The lookup we used:

The "misery index" for a player is the sum of the three binary classifiers — integer from 0 (not a bust by any definition) to 3 (triple bust):

The "triple-bust rate" of a franchise is the proportion of mature picks (i.e., picks made at least three years ago) for which misery_index = 3.

Why intersection over union? A union-based definition (any of A, B, or C is true → "bust") would label most NFL first-rounders as busts and lose discriminative power. The intersection forces agreement across three lenses, producing a number that survives skepticism from any direction.

Run it yourself, on any team, in any year

The full pipeline (data acquisition + classification + summary statistics) is one Python file. The methodological core — the part that takes a player record and returns the three classifications — is below:

# Expected career AV by first-round slot. From Chase Stuart's published
# expected-AV-per-pick curve. Used as the slot-median proxy for Definition B.
EXPECTED_AV = {
    1: 72,  2: 66,  3: 60,  4: 54,  5: 49,  6: 45,  7: 41,  8: 38,
    9: 35, 10: 33, 11: 30, 12: 28, 13: 27, 14: 25, 15: 24, 16: 22,
    17: 21, 18: 20, 19: 19, 20: 18, 21: 17, 22: 16, 23: 15, 24: 14,
    25: 13, 26: 13, 27: 12, 28: 11, 29: 11, 30: 10, 31: 10, 32:  9,
}

# Cutoff: only picks made at least three years ago are "mature" enough to
# judge. A player drafted in 2024 has not had a fair chance to make four
# Pro Bowls or four starter-year tenures.
MATURE_CUTOFF = 2022

def classify(pick):
    """Return the three bust flags + the misery index for one pick."""
    year, slot, name, pos, pb, av, styr, notes = pick
    exp_av = EXPECTED_AV.get(slot, 20)

    if year > MATURE_CUTOFF:
        return {"bust_A": None, "bust_B": None, "bust_C": None,
                "misery": None, "expected_av": exp_av}

    return {
        "bust_A": pb == 0,                    # never made a Pro Bowl
        "bust_B": av < exp_av,                # below slot's expected AV
        "bust_C": styr < 4,                  # under 4 NYJ starter years
        "expected_av": exp_av,
        "misery": sum([pb == 0, av < exp_av, styr < 4]),
    }

To reproduce Issue #25's numbers: clone the repo, run python3 scripts/build_jets_draft_dataset.py, and the script writes scripts/data/jets-first-round-1976-2025.csv with one row per pick and a misery_index column.

To run it on a different team: replace the PICKS list with that team's first-round data (same eight-tuple format: year, slot, name, position, career Pro Bowls, career AV, primary-starter seasons for the drafting team, notes). Everything else carries over.

To run it across all 32 teams: the upcoming “After the Jets” series (May 2026) will publish that dataset and a generalized version of this script. Watch for Issues #31 through #40 in the queue.

Caveats and known limitations

• Career AV figures are from Pro-Football-Reference and reflect totals as of early 2026; players still active will accumulate more.
• The Expected AV curve is a published average; individual slot-years vary considerably. Treat it as a benchmark, not a precise prediction.
• Definitions A, B, C have judgment calls baked in (4 years for "primary starter," 0 Pro Bowls as the cutoff). Reasonable people could move those thresholds and re-run.
• Dataset compilation involved manual recall and spot-verification for borderline cases. The CSV is the authoritative record; if any specific player line conflicts with the canonical PFR record, file a GitHub issue and we'll fix it.