How we compute this

Tickets printed ≈ prizes in a tier × that tier's printed odds; we take the median across tiers to damp rounding noise. Lotteries publish exactly two per-tier facts (printed odds and remaining-of-total counts), and everything below derives from those.

% sold: tiers with thousands of prizes are a decent random sample of the print run, so remaining ÷ total over high-count tiers estimates the share of tickets still on shelves. Low-count tiers are excluded — five top prizes tell you nothing about sales.

EV per $1 = prize money still unclaimed ÷ estimated tickets still unsold ÷ ticket price. It is always below $1 — the house keeps the rest. Annuity top prizes are counted at ≈60% of face value (rough 25-year cash equivalent), not the advertised total.

Luck z: under a uniform-shuffle null, top prizes remaining ~ Binomial(n, fraction unsold). z measures how far the observed count sits from that expectation. With a handful of top prizes the raw remaining/expected ratio is mostly noise, so badges and the score key on z (|z| ≥ 1.5), never the ratio; beyond |z| = 3 the null itself is broken, so we report a data anomaly, not luck.

Score (0–100): EV dominates; statistically significant luck adds or subtracts a little; boards with no top prize left, or with $10k+ tiers depleted well below what sales predict, sink hard.

Caveat: states post claims with a lag of days to weeks, so every count here trails reality. A "hot" board may already have been hit.

Read the full methodology →