The winner-take-most shelf

Claim (verbatim)

Survival among the early Persian poets is not merely low, it is radically unequal, and the shape of that inequality is a fingerprint of the filter that produced it. Take every poet Ganjoor assigns to the period before the Mongol conquest and rank them by the size of their surviving corpus. If loss had been an even tax on all - a fixed survival probability per line - the survivors would still roughly preserve the relative sizes of the original outputs, and no single poet would tower absurdly over the rest. But survival was not a per-line lottery; it was decided by canonization. A poet who became a school-text, an anthology staple, a name a lexicographer reached for, kept a whole divan; his equally-productive contemporary who missed that gate kept a few quoted bayts or nothing. Canonization is a winner-take-most process, so the surviving-corpus distribution should be brutally top-heavy - a handful of names holding almost all the text, a long tail of near-empty entries - far more skewed than any plausible distribution of original output. The concentration is the loss made visible: it measures not who wrote the most, but who got through the gate. Prediction restated: the top ten pre-Mongol poets on Ganjoor hold the overwhelming majority of all pre-Mongol surviving verse, and the median such poet is left with almost nothing.

Prediction clause (verbatim)

Prediction: among Ganjoor poets with a death-date before 1220, the ten largest surviving corpora will together account for over 80 percent of all pre-1220 surviving bayts, while the median pre-1220 poet holds under 200 bayts (primary clause: the top-ten share above 80 percent of pre-1220 bayts; the verdict follows it). Poets are assigned to the pre-Mongol set by death-dates from Safa and de Blois, the count is over attributed poem-text on Ganjoor, and the test voids if fewer than 40 pre-1220 poets carry a datable identity.

Kill-dataset (verbatim)

Kill: the Ganjoor (ganjoor.net) author roster with per-poet bayt totals, restricted to poets dated before 1220 via Safa's Tarikh-i adabiyat dar Iran and de Blois; compute the top-ten share of total bayts and the median per-poet corpus. A within-corpus concentration operation, no external attestation needed.

Novelty / leakage triage

anticipated in the literature — this exact test has never been run

That a handful of canonized names dominate pre-Mongol survival while most attested poets are reduced to fragments or nothing is the qualitative burden of Lazard's fragment collection and of the survey literature's per-poet inventories. But a concentration statistic - the top-ten share of pre-1220 bayts on a digital corpus, plus a median-poet figure - has never been published, and both numbers depend non-trivially on Ganjoor's roster composition (how long and how thin its pre-1220 tail actually is), so the answer is not guaranteed by anything in print.

• F. de Blois, Persian Literature: A Bio-Bibliographical Survey, vol. V (2nd rev. ed., London, 2004), per-poet inventories of the surviving corpora
• G. Lazard, Les premiers poetes persans (IXe-Xe siecles), 2 vols (Tehran-Paris, 1964)
• Z. Safa, Tarikh-i adabiyat dar Iran, vols I-II

Predictions

No prediction registered yet.