Equal odds, different choices
Powerball has 292,201,338 valid jackpot combinations under the current rules: choose five of 69 main numbers, then one of 26 Powerballs. Every valid combination has the same jackpot probability. Our analysis does not change it.
Generation and indicators
The free builder uses cryptographic random integers with rejection sampling. Main numbers are selected without replacement after preserving your chosen numbers. Above-31 count, odd/even split, sum, span and adjacent consecutive pairs are calculated directly from the line. No prediction model is used.
Historical data and scope
Draw records come from the New York State Gaming Commission dataset. We retain draws from October 7, 2015 onward to avoid mixing incompatible number ranges. The results page shows the import timestamp and links to the source. A scheduled import validates records before publishing them.
Historical replay
Labs compares a supplied line with every draw in the selected window. It reports all prize-tier matches and the number of non-matching draws. It is retrospective: the line was not necessarily generated before those draws. Monetary prizes are not calculated, because rules and payouts can vary.
SplitGuard model and limits
SplitGuard v4 fits a regularized Poisson count model to official three-main winner counts: each draw has thousands of these winners, compared with a handful of jackpot-tier winners, so number-pattern popularity can be measured reliably. Exposure is inferred from the 2+1, 1+1 and 0+1 tiers; it is not reported ticket sales. The model was chosen in a challenge whose candidates and rules were fixed in advance: five chronological development folds (61.8% deviance improvement on all data, every fold improving) and an untouched holdout of 351 draws from June 2024 to September 2026 (57.1% improvement). On that holdout the same pattern effects also predicted jackpot-tier winner counts slightly better than no model (+1.2%): a small but non-negative gain, because jackpot-tier winners are rare. The index is normalized to a mean of 100 over 65,536 fixed uniform reference lines. Powerball effects are neutral because they have not been validated. Alternatives are calculated exhaustively; every returned line has a strictly lower unrounded score than the original line. The estimate describes broad player-choice patterns: it cannot count exact matching tickets or guarantee fewer co-winners.
Powerball official prize-tier results ↗ · NY State draw-number dataset ↗