NumbersIntel is an independent lottery-analytics site. We don't predict winning numbers — nobody can. We compute transparent, reproducible statistics from the official draw record so you can see what has actually happened. This page documents every source, window and formula behind the figures.
Where the data comes from
Winning numbers, jackpots and cash values come from official lottery sources — each game's state or national lottery, or the multi-state operators (powerball.com / Mega Millions' data service) for the national games, supplemented by official state results and well-established public archives where an official feed isn't machine-readable. Every game's data file records its source and an earliest-available date. We currently track roughly 150 draw games across 47 US jurisdictions (all 50 states that run a lottery, plus Washington D.C. and Puerto Rico) and the UK National Lottery.
How often it updates
An automated pipeline (a Python scraper run on a GitHub Actions schedule) refreshes results
after each drawing and commits the new data, which deploys automatically. Each game page shows
the date of its latest draw, and our sitemaps carry real lastmod timestamps so
search engines see the freshness. When a source changes or a scrape fails, the pipeline keeps
the last known-good values rather than publishing blanks.
Number-frequency statistics
For each game we count how often every number has been drawn across the available history, then derive:
- Most / least common numbers — raw draw counts over the selected window (all-time by default; date-range filters narrow it).
- Hot & cold numbers — the most- and least-drawn numbers within the chosen window. "Hot" simply means drawn more often recently; it carries no predictive power.
- Overdue numbers — the number of draws since each number last appeared (its current gap). The "most overdue" number is the one with the longest current gap.
- Frequency by pool — main balls and any special ball (Powerball, Mega Ball, Lucky Stars, Cash Ball, etc.) are counted separately, since they're drawn from different pools.
Digit games (Pick 3/4/5-style) are analysed by digit (0–9), and the same windows apply.
Odds
Jackpot and prize-tier odds are computed combinatorially from each game's published matrix — e.g. a 5-of-69 + 1-of-26 game has jackpot odds of C(69,5) × 26 = 1 in 292,201,338. We show the exact matrix on each game page. Odds are mathematical constants of the game; they don't change from draw to draw, and no number, hot or cold, is more or less likely than any other.
Expected value (where shown)
For games with a fully published, fixed prize structure we compute value per $1 spent: the cash-option jackpot plus every fixed lower-tier prize, each divided by its odds, summed, multiplied by an assumed 63% post-tax take (about 37% federal plus a blended state estimate), and divided by the ticket price. The result is in cents returned per dollar. Games whose lower tiers are pari-mutuel (they vary every draw) are shown with odds only, not a single value-per-dollar, because a fixed EV would be invented. UK games are shown in pounds and are excluded from the US-tax value rankings.
Value vs Heat: the two dials
The Value vs Heat page runs two indices that are deliberately never combined into a single score, because they measure different things and routinely disagree.
Value is the expected return per dollar wagered, adjusted for the chance of sharing the jackpot. For each draw:
- The jackpot is taken at its cash value, never the advertised annuity. The advertised figure is a stream of payments whose relationship to cash moves with interest rates — the ratio ran roughly 0.72 in 2021 and about 0.50 by 2024 — so the same headline is worth materially different amounts in different years.
- It is multiplied by the expected share a winner keeps, (1 − e−λ) / λ, where λ is estimated tickets in play divided by the jackpot odds. This is the exact expectation of 1/(1+K) when the number of other winners K is Poisson-distributed.
- Every fixed lower tier's expected value is added, the total divided by the ticket price, and the same 63% post-tax factor applied as elsewhere on this site.
- The result is reported as a percentile against that game's own history, because a raw figure in cents is hard to place. "6th percentile" says more than "12.3¢".
Estimated tickets are derived, not assumed. For every past draw we invert the lottery's published per-tier winner counts against those tiers' odds — if 700,000 tickets matched a tier whose odds are 1 in 38, roughly 27 million tickets were in play. Pooling all tiers gives a maximum-likelihood estimate of participation. The relationship between jackpot size and tickets sold is then read from the game's own record, not modelled.
Heat is a separate attention index from three parts, each converted to a percentile and averaged: consecutive rollovers, sales velocity against the game's trailing median, and advertised-jackpot percentile. It is a measure of mania, not of worth.
Value is not increasing in jackpot size. Past a peak, a growing jackpot sells tickets faster than it grows, so the expected share falls faster than the prize rises and value per dollar declines while the headline climbs. On Powerball's record since 2015 that turn currently sits near $1.8bn advertised. The shape is robust; the exact turning point rests on very few observations, since only a handful of draws have ever passed $1bn.
Percentiles are matrix-aware. A prize-matrix change makes earlier draws a different game, so each is ranked only within its current matrix — Powerball from its October 2015 change, Mega Millions from its April 2025 relaunch to a $5 ticket with a built-in multiplier.
What we do not compute. The cross-game leaderboard includes only games whose prize structure is published in full. Games with pari-mutuel tiers, whose prizes depend on sales we cannot observe, are omitted rather than estimated. Games with published fixed tiers but an unpublished rolling jackpot are shown as a lower bound and labelled "at least". Where per-tier winner counts lag the live jackpot and the jackpot has since been won, the rollover count is reported as unknown rather than as a stale number.
One figure deserves a warning. Before tax, split-adjusted value per dollar genuinely exceeds $1 at extreme jackpots — our own sweep peaks near 142¢ per dollar. That is a real, long-documented result, but it is not an argument for buying a ticket: winnings are taxable, the advertised prize is an annuity, and expected value says nothing about a roughly 1-in-292-million chance. We therefore display the post-tax figure.
What these statistics are not
Lottery draws are independent random events. Frequency, hot/cold and overdue statistics are descriptive — they summarise the past. They do not predict future draws, and no pattern in past results changes the odds of the next one. We publish these figures because they're interesting and verifiable, not because they offer an edge. There is no system that beats a game of chance.
Corrections & transparency
Every game's underlying data file is downloadable as JSON, linked from its results page, so anyone can audit our numbers. If you spot an error, email contact@numbersintel.com and we'll fix it.
Responsible play: the lottery is a negative-expected-value game by design — for entertainment only, never an investment. You must be 18+ (21+ in some states). If gambling is a problem for you or someone you know, call 1-800-GAMBLER (1-800-426-2537). This site is information and entertainment only, not financial, legal or gambling advice.