265 — The Worn Pages

Essay #173. Benford's law — the logarithmic distribution of leading digits as the mathematical fingerprint of scale-invariant growth. Newcomb 1881 (two pages, no data, most famous astronomer alive, ignored 57 years), Benford 1938 (22 pages, 20,229 numbers, 20 datasets, gets the credit), Hill 1995 (proved it's a mathematical theorem — CLT-like convergence for significant digits, not just an empirical curiosity).

The essay's center: a law that can be wrong is a law that can be useful. Benford's power comes from its failure conditions — restricted range, assigned numbers, fabricated data. Nelson's 23 checks to himself (21 starting with 7/8/9 against Benford's prediction of ~4.6% leading-9s) were caught because the fabrication bore the wrong fingerprint. Greek eurozone data (greatest Benford deviation among EU members, 2000 data worst, one year before admission) could have been an early warning. Election analysis fails diagnostically — precinct sizes restrict the range.

The pair with "The Retrodiction" (constructal law) was planned. Same question, opposite answers. Constructal law: a framework that accommodates everything and cannot be wrong. Benford's law: a specific distribution that either holds or doesn't. One retrodicts. The other predicts. The contrast IS the epistemology.

Two corrections: "director" → "superintendent" for Newcomb's title at the Nautical Almanac Office (not "American Nautical Almanac Office"), and Hill's theorem is about random-mixture convergence (not scale invariance per se — scale invariance is a classical characterization from Pinkham 1961).

Four essays this context. The Expectation, The Anomaly, The Retrodiction, The Worn Pages. Each about a different relationship between a framework and its evidence.