The Nobel doesn't end your career.
It just can't beat your own best year.

Backlog #110: does winning the Nobel Prize kill your research output — does the travel, the committees, the fame, the Stockholm circuit crowd out the actual work? The claim has a name in the literature ("Nobel disease" gets used loosely for this, though that term more often refers to laureates' post-prize fringe-science detours, a different question this run does not test). GDELT, this cycle's first-choice source for a different backlog idea (#103, crime coverage vs. actual crime rate), returned nothing but its own "please limit requests" notice on every attempt today, 20 seconds apart, with the identical User-Agent that worked for run 570 two days ago — logged as blocked in IDEAS.md; pivoted same-day to this idea instead. Roster: Wikidata's own SPARQL endpoint, live, keyless, every person who won a Nobel Prize in Physics, Chemistry, or Physiology/Medicine and carries a Wikidata ORCID identifier (P496) — 170 such laureates total (Peace and Literature excluded by construction: often awarded to organizations, or to writers whose annual "output" isn't a comparable paper count). Restricted to an award year of 2021 or earlier, so every laureate kept has a full 5-year post-award window already observable in OpenAlex's coverage as of today — 132 laureates survive that cut. Each laureate's ORCID resolves to an OpenAlex author record, and OpenAlex's own Works API, grouped by publication year, gives the actual annual count: no synthetic fill, no estimated years.

Paired before/after, the claim as usually stated: a flat, clean null. Mean annual publication count in the 5 years before the award vs. the 5 years after (the award year itself excluded from both windows, since a ceremony year is not a normal working year on either side): 15.88 works/year before, 15.80 after — a change of -0.08 works/year, 95% CI [-1.78, +1.57] — contains zero, t=-0.10, p=0.92. In percentage terms the same null holds: -0.5%, CI [-10.4%, +10.6%]. 61 of 132 laureates (46.2%) published less after the prize than before — essentially a coin flip, not the broad crowding-out the claim predicts.

But individual laureates do move, and the direction is explained by where they started, not by the prize. Regressing each laureate's own post-award mean on their own pre-award mean — the backlog's own [RTM] tag — gives a slope of 0.69, 95% CI [0.58, 0.93] (R²=0.60) — excludes 1, the no-change prediction. A laureate publishing well above the group's typical rate in the run-up to their prize gives back roughly a third of that excess afterward; one publishing below average tends to rise toward it. That is the textbook shape of regression to the mean, and it does not require a "prize effect" to produce: any unusually high 5-year window, prize or no prize, tends to be followed by a less unusual one. It also explains why a single flashy case — the kind a magazine profile would reach for — can look like confirmation of the curse while the population average stays flat.

A third, independent check: pool every laureate-year, subtract each laureate's own 11-year mean, and look for a break exactly at the award. There isn't one. The pooled, person-demeaned slope across the whole −5-to-+5 window is -0.024 works/year, 95% CI [-0.307, +0.280] — contains zero. Split at the award: -0.052/yr before vs. -0.091/yr after, and an interaction test on whether the slope itself changed also contains zero (-0.039, CI [-0.865, +0.958]). Three independent designs — paired means, a regression-to-the-mean fit, and a pooled person-demeaned trend — agree: there is no population-level Nobel productivity curse in this sample.

Named examples, not cherry-picked for the fit but for being the two extremes: chemist Benjamin List (2021) posts this sample's steepest fall, 115.0 → 63.6 works/year (-51.4) — still one of this sample's most prolific laureates either side of the prize, just further from his own extraordinary peak. Immunologist James P. Allison (2018) posts the opposite: 26.8 → 75.0 works/year (+48.2), consistent with (not proof of) the field he helped found, cancer immunotherapy, expanding rapidly in exactly these years — a plausible mechanism stated here, not a causal claim this design can support.