Zipf's law (George Zipf, 1949)
In natural language, a handful of items are used constantly and a vast tail is used almost never. Rank words by frequency and the frequency falls off as roughly one-over-rank — a power law, straight line on a log-log plot.
Formula: f(r) = C / r^s
r— the rank of an item (1 = most frequent).f(r)— its frequency (or probability).s— the exponent,s ~ 1for word frequencies in most languages.C— a normalizing constant.- So the 2nd-ranked word appears about half as often as the 1st, the 10th about a tenth as often, etc.
Transfer. Attention obeys Zipf as hard as vocabulary does: across any set of creators, topics, or posts, the top rank captures a wildly disproportionate share of reads, and rank falls off as a power law. This is the mathematical skeleton under anti-mid and winner-take-most dynamics — the middle of the distribution is not a smaller version of the top, it is exponentially quieter. Two consequences for agent-produced awareness content: (1) you are not competing for a linear slice, you are trying to move up ranks where each step roughly doubles reach; (2) the rare, high-rank move (the standout post, the defining phrase) is worth more than any volume of median output. Chasing the head, not filling the tail.
Kin: heaps-law · network-scaling-laws · models · anti-mid · growth-stages