Method · Explanation Burden

How much explanation does one event require?

Explanation Burden measures explanatory difficulty. It does not measure suffering, event magnitude, team loyalty, or whether an event qualifies for the Archive. Annual adjudication still decides what enters the record.

Active axis · 0–5

Mechanism Complexity

Nonredundant causal dependencies that must be held together to explain the event.

0: One ordinary cause fully explains it
1: Two linked facts, one system
2: Three dependencies, or two interacting systems
3: Four or five dependencies across multiple systems
4: Six or more, multiple stages, feedback, or delayed consequences
5: Recursive or self-negating — an attempted correction disables or reverses itself

Active axis · 0–5

Closure Deficit

Explanatory resistance remaining after the best ordinary explanation.

0: A common mistake or ordinary cause fully closes it
1: Complete and intuitive
2: Complete but unusual
3: Material facts disputed, incomplete, or inconsistent
4: The official or technically correct explanation remains strongly counterintuitive
5: Every component may be correct, yet the aggregate remains recursive or self-defeating

Total 0–10

Five bands. Profile first.

No weighting. The total is Mechanism plus Closure. Ties are acceptable. The archive does not publish a 1-through-18 league table.

0–2: Ordinary
3–4: Elevated
5–6: Anomalous
7–8: Severe
9–10: Extreme

Display format example: 9/10 — Extreme · Mechanism 4 · Closure 5

Withdrawal record

Rarity Evidence is not an active axis.

Removed 2026-08-11 after Pilot A. Blind raters did not share a convergent interpretation. Zeros assigned on that axis were 19, 1, and 18 out of 30.

It does not appear as a value, field, score, or recoverable option on this instrument. The index is two axes and is scored out of ten.

Pilot A · adopted medians

Fifteen scores. Three still blank.

Median of three independent blind raters, 2026-08-11. The fifteen adopted medians were independently reproduced from recovered local-session rating tables, original analysis code, and checksummed packet records. Substantial recovery verifies the mathematical calculation and permutation result (p = 0.0099), while provider-native response and session transcripts, exact provider model versions and provider session identifiers, and server-side packet-delivery and randomization-generation records remain unpreserved.

Fifteen adopted Pilot A medians
Case	Year	Title	Mechanism	Closure	Total	Band
Case 005	2012	The Unreviewable Automatic Review	5	5	10/10 — Extreme · Mechanism 5 · Closure 510/10	Extreme
Case 013	2020	The Unwanted Touchdown	5	5	10/10 — Extreme · Mechanism 5 · Closure 510/10	Extreme
Case 002	2009	The Cleveland Timeout	4	5	9/10 — Extreme · Mechanism 4 · Closure 59/10	Extreme
Case 010	2017	Eight Seconds Remaining	4	5	9/10 — Extreme · Mechanism 4 · Closure 59/10	Extreme
Case 015	2022	The Delayed-Action Violation	4	3	7/10 — Severe · Mechanism 4 · Closure 37/10	Severe
Case 003	2010	The Process of the Catch	2	4	6/10 — Anomalous · Mechanism 2 · Closure 46/10	Anomalous
Case 016	2023	Who Reported?	3	3	6/10 — Anomalous · Mechanism 3 · Closure 36/10	Anomalous
Case 017	2024	The Defense of Theseus	3	3	6/10 — Anomalous · Mechanism 3 · Closure 36/10	Anomalous
Case 014	2021	Tucker II: Subjective Zero	2	2	4/10 — Elevated · Mechanism 2 · Closure 24/10	Elevated
Case 006	2013	Tucker I: 61	2	1	3/10 — Elevated · Mechanism 2 · Closure 13/10	Elevated
Case 007	2014	The Flag That Wasn’t	2	0	2/10 — Ordinary · Mechanism 2 · Closure 02/10	Ordinary
Case 008	2015	The Illegal Bat	2	0	2/10 — Ordinary · Mechanism 2 · Closure 02/10	Ordinary
Case 009	2016	The Facemask Catch	2	0	2/10 — Ordinary · Mechanism 2 · Closure 02/10	Ordinary
Case 012	2019	Hands to Somewhere Near the Face	2	0	2/10 — Ordinary · Mechanism 2 · Closure 02/10	Ordinary
Case 001	2008	Terminal Zeroing: 0–16	0	1	1/10 — Ordinary · Mechanism 0 · Closure 11/10	Ordinary

Unscored

Case 004 · 2011

The Whistle Before the Recovery

Unscored — not included in Pilot A.

Not represented in Pilot A.

Case 011 · 2018

They Knew the Plays

Unscored — not included in Pilot A.

Not represented in Pilot A.

Case 018 · 2025

Two Touchdowns, Zero Points

Unscored — not included in Pilot A.

Verified after Pilot A. Adopted into the Archive; excluded from Pilot A diagnostics.

What held

Case 001 sits at the floor. That is correct.

The terminal zeroing is ordinary failure at maximum magnitude. A bad team lost every game, which is one sentence. Under the 0–16 Turn its significance is positional, not explanatory, and Archive Role carries that weight.

The four plain-error cases — 2014, 2015, 2016, 2019 — all scored 2/10, from raters who were never told the closure axis was inverted. A plain officiating error is a complete explanation. That is the strongest validation the instrument has.

Allowed result

Harder to explain. Not rarer.

Archive events scored a median of 6.0/10. Pre-2008 events scored a median of 3.0/10. A permutation test returned p = 0.0099.

A non-Detroit control scored 6.0/10. The two most painful events in the set scored 1.0/10 each. The instrument measures explanatory difficulty, not suffering and not Detroit.

The archive may claim that post-2008 events are measurably harder to explain than pre-2008 events. It may not claim that they are rarer. Separation came from mechanism.

The curse did not become less likely to be explained. It became harder to explain.

Limitations

Published with the result.

n = 30. A pilot, not a study.
Three events are recognisable to anyone who follows the sport.
The raters had prior project exposure. Blinding the events is the control.
Sampling is asymmetric between the two arms. The true separation may be smaller.
Event descriptions were written by one author who knew the hypothesis.
Cases 004, 011, and 018 remain unscored.
The original scoring packet, answer key, three complete session rating tables, analysis code, seed, and permutation outputs have been recovered and exactly reproduce the fifteen adopted medians and study p-value. Provider-native response and session transcripts, exact provider model versions and provider session identifiers, and server-side packet-delivery and randomization-generation records remain unpreserved as documented limitations.