The Lindy effectNassim Taleb's rule: the longer something non-perishable has been around — a book, a technology, a company, an idea — the longer it is likely to keep going. Applies to things that don't age out (a person's life expectancy, by contrast, decreases with age). says a non-perishable thing's future life is proportional to its past — a book in print 50 years probably has 50 more in it; one that came out last year, probably not 50. Under a power-law tail with exponent α, the odds of surviving another r is (a / (a + r))αP(remaining > r | survived to a) = (a / (a + r))^α. At α = 1 (strict Lindy / Copernican), the median remaining lifetime equals the current age..
Even odds it lasts at least 30.0 years longer (and reaches age 60.0 years). The upper quartile reaches 90.0 years; the long-tail 90th, 270 years.
The dashed crosshair is the median: a 50/50 chance it lasts at least that much longer. The curve always passes through 50% at r = a · (21/α − 1); at α = 1 that's exactly the current age.
| It survives at least | Additional years |
|---|---|
almost certainly survives (10%) survival ≥ 90% | 3.33 years |
lower quartile (25%) survival ≥ 75% | 10.00 years |
median — even odds (50%) survival ≥ 50% | 30.0 years |
upper quartile (75%) survival ≥ 25% | 90.0 years |
long tail (90%) survival ≥ 10% | 270 years |
Read each row as "with this much confidence, it lasts at least this long" — the 25% row is the safest bet (almost certainly true), the 90% row the optimistic tail. Pair this with the rule of three (what a clean run can promise) and the regression to the mean for two more cuts at "what does past behaviour really predict".