Half the poets are a single verse

Claim (verbatim)

Vidyakara's Subhasitaratnakosa, compiled at the Jagaddala monastery in Pala Bengal around 1100-1130 and edited by D. D. Kosambi and V. V. Gokhale as Harvard Oriental Series 42 (1957), preserves roughly 1,738 verses under the names of over two hundred poets, and its distribution of verses-per-poet is not flat but savagely long-tailed: a handful of canonical names carry dozens of verses each while the great majority of named poets appear exactly once. That shape is the fingerprint of a drowned literature sampled almost at random. An anthologist copying from what happened to be circulating will catch the famous many times over and the obscure once, so the single-verse poets, the hapax kavayah, are the visible edge of a population whose works are otherwise entirely gone. The Harvard edition's kavi-index makes the tail countable, and Ludwik Sternbach's Maha-subhasita-samgraha concordance lets one pool a poet's total surviving footprint across every major anthology at once. The claim is that for most named poets the anthology simply is the collected works: their entire literary afterlife is one or two quotable stanzas, with nothing under their name surviving as an independent text anywhere in the manuscript record. The long tail is not a quirk of taste but the exposed cross-section of a lost majority.

Prediction clause (verbatim)

Prediction: in Vidyakara's Subhasitaratnakosa more than half of the distinct named poets will be represented by exactly one verse (single-verse hapax poets), and, pooling each such poet's verses across the major printed anthologies via Sternbach's Maha-subhasita-samgraha, the median total surviving footprint of the hapax poets will be no more than two verses against zero independently transmitted work; the greater-than-one-half single-verse-poet fraction is the deciding clause and the footprint figure follows it. Attribution is scored on the edition's own kavi-index, verses of uncertain or variant attribution are excluded before counting, and if fewer than 150 distinct attributed poets resolve the test voids for coverage rather than resolving.

Kill-dataset (verbatim)

Kill (not yet built): the kavi-index of Kosambi and Gokhale's Subhasitaratnakosa (Harvard Oriental Series 42) for the per-poet verse counts, cross-checked against the SARIT e-text of the anthology for the attributions, with the pooled footprint taken from Sternbach's Maha-subhasita-samgraha concordance and independent survival checked against the New Catalogus Catalogorum.

Novelty / leakage triage

already answered in the literature

The per-poet verse counts are already a printed tabulation: the kavi-index of Kosambi and Gokhale's Harvard edition lists every attributed poet with his verse-numbers, and Kosambi's introduction itself runs a statistical analysis of the poet roster. Ingalls' companion volume states the qualitative answer outright: the anthology quotes over two hundred poets, most represented by only one or two stanzas and otherwise unknown. The item's pooling instrument also already exists as a published book, since Sternbach's Descriptive Catalogue is precisely the per-poet footprint across the anthologies with the otherwise-unattested poets flagged. Only the final percentage read-off is left, which is share-arithmetic on published tables, not an open measurement.

• D. D. Kosambi and V. V. Gokhale (eds.), The Subhasitaratnakosa Compiled by Vidyakara, Harvard Oriental Series 42 (Cambridge, Mass., 1957), Kosambi's introduction and the kavi-index
• D. H. H. Ingalls (trans.), An Anthology of Sanskrit Court Poetry: Vidyakara's Subhasitaratnakosa, Harvard Oriental Series 44 (Cambridge, Mass., 1965), general introduction and verse-notes
• L. Sternbach, A Descriptive Catalogue of Poets Quoted in Sanskrit Anthologies and Inscriptions, 2 vols. (Wiesbaden, 1978-1980)

Predictions

No prediction registered yet.