Predictions — The Numbers Lab

The founding verdict

Before this section opened, the lab pre-registered a test of the strategies a numbers magazine usually sells — follow the hot numbers, play the due ones — and ran it on the official New York history. Thirteen strategy cells, 8,652 bets per cell, one $1 straight bet per draw from 2015 through November 2026. Result: no edge. No cell reached statistical significance (smallest p = 0.22 against a pre-registered bar of 0.01). Pooled across the twelve straight-bet cells: 51 hits against 51.9 expected from chance alone, a pooled return of about −49% — the house edge, exactly as independence predicts.

That verdict is the reason this page publishes no picks. It is also the standing test protocol: any future method proposed for this section is pre-registered first, run against the full fetched history, and published whatever it shows.

The only terms picks may ever publish on

If picks are ever published here, they publish under all of these terms, stated on the page with them:

1. Each pick appears before its draw, timestamped, and enters the picks ledger permanently — wins, losses, and the base-rate line beside them.

2. Each pick is labeled entertainment, with the standing line: "No method improves the odds of this game. These picks are tracked for the record and are expected to return about 50 cents per dollar over time, like any other selection." The expected return is not fine print: across these games, about 50 cents of every $1 comes back in prizes over time, whichever numbers are chosen.

3. The selection rule for a pick series is stated in full on the page and frozen while the series runs. A series is never renamed, restarted, or quietly dropped.

4. No pick is ever described with performance language its ledger does not support, and no consensus of entertainment rules is ever called an edge.

Descriptive statistics — past draws only. Every draw is independent: every combination has the same probability on the next draw, no matter what these tables show.