Does This Look Random to You?

THE RANDOMNESS DOCKET

Does This Look Random to You?

Real Powerball drawings can make ordinary randomness look suspicious. This piece treats three patterns as exhibits: consecutive numbers, repeated numbers between drawings, and a number that seems overdue. Each one is compared with what a fair random process predicts.


Exhibit A. A single Powerball drawing.

Exhibit showing five Powerball balls from the February 20, 2019 drawing — 27, 49, 50, 51, 52 — with the observed rate of consecutive pairs (26.4%) compared to a fair-random simulation (26.3%).

Exhibit B. Two consecutive drawings.

Exhibit showing two consecutive Powerball drawings, February 9 and February 11, 2026, sharing the numbers 6 and 48, with the observed rate of at least one shared number (32.0%) compared to the hypergeometric expectation (32.2%).

Exhibit C. A number that hasn’t appeared in weeks.

Exhibit showing number 23's current 48-draw drought compared to the typical range of 46 to 87 draws for the most-overdue number, from a 5,000-replicate simulation.

MECHANISM

Three exhibits, three intuition failures. Here’s the property of randomness that’s doing the work behind all of them.

No number in this dataset appears more or less often than chance would predict. But that stability only shows up over the long run. Early in any single number’s history, its observed rate swings 15 to 300 times more than it does over the full dataset — small samples are noisy by nature. As draws accumulate, that swing narrows toward the expected rate.

A consecutive pair, a shared number, a long drought — each looks like a signal because it’s being read in isolation, the way a single early swing looks like a trend before the sample catches up. Looked at across the full history, each is part of the texture a fair random process produces at this sample size.


Randomness isn’t the absence of patterns. It’s the presence of patterns that don’t mean anything.


Data source. NY State Gaming Commission, Powerball Winning Numbers, 2010–present. Current-era statistics (Panels 1–3) use 2015–present draws only, matching the 2015 white-ball pool-size change.

Tools. R, tidyverse, ggplot2. Simulation: 5,000-replicate Monte Carlo, real draw mechanics (5-of-69, no replacement), seed 20260702.