We Analyzed 2,700+ Scratch-Offs Across 42 U.S. Lotteries

Why an engineer scraped 42 lotteries

In 2003, a retired factory manager named Jerry Selbee read the rules of a Michigan lottery game and noticed the math briefly worked in the player’s favor. He and his wife Marge made millions before the game closed, and they broke no rules. They just did arithmetic nobody else bothered to do.

I’m an aerospace engineer, and that story never left me alone. Scratch-offs publish their arithmetic too. Most state lotteries post, on their own websites, how many prizes at each level are still unclaimed in every active game. Almost nobody reads it. So I built a system that reads all of it, every day, across 41 states and Washington, D.C. This article is what that dataset says, in aggregate, as of July 12, 2026.

The numbers below cover 2,706 active scratch-off games and 2.38 million individual prize-tier records, collected every day since March 17, 2026. Every figure traces back to public lottery data, and the math we run on it is documented on our methodology page.

Finding 1: the average ticket returns about 69 cents per dollar

Across the 2,687 active games with published expected-value data (of 2,706 tracked), the average expected return is $0.69 for every $1 spent. The median game is nearly identical at $0.68. If the phrase “expected return” is new, our EV explainer walks through it, but the plain version is: for each dollar that goes into an average scratch-off, about 69 cents comes back out as prizes, spread across all players. The other 31 cents is the lottery’s.

That is not a scandal. It is how the products are filed and designed, and it is roughly in line with the payout percentages state lotteries publish for their scratch portfolios. What the averages hide is how much the individual games differ, which is where the rest of the findings come in.

Finding 2: the cheapest tickets are the worst deal

The clearest pattern in the entire dataset: expected return climbs with ticket price, at every step of the ladder.

A $1 ticket keeps about 42 cents of your dollar. A $50 ticket keeps about 20. The gap is not an accident: lotteries design higher price points with richer prize pools to justify the price tag, and the design shows up plainly in the published numbers. Sold by the dollar, the cheapest rack position is consistently the most expensive place to stand.

Finding 3: about 1 in 9 games on sale can no longer pay its top prize

As of July 12, 2026, 294 of the 2,704 active games with published prize-tier data (two of the 2,706 tracked games lack it), about 10.9 percent, show zero top prizes remaining. The jackpot printed in big numbers on the front of the ticket has already been claimed by someone else. The tickets stay on sale, and in most states that is legal and disclosed, if you know which webpage to check.

This is the single most practical thing in the dataset. Two tickets can sit side by side in the same case, same price, same art style, and one of them still has its million-dollar prize in circulation while the other physically cannot produce one. The only way to tell them apart is the remaining-prize table, which is exactly the public data most players never see. Florida, for example, publishes it for every game on the state lottery’s own site. A few lotteries publish less granular data whose tier counts we model from published odds (documented on the methodology page); excluding those states barely moves this figure. It stays right around 1 in 9.

Finding 4: your state matters more than you think

The median game in the strongest state returns about 77 cents per dollar (Massachusetts, across 133 active games). The median game in the weakest returns about 53 cents (Kansas, across 62 games). That is a 24-cent spread on every dollar, purely as a function of which side of a state line the gas station sits on.

The spread is a snapshot and it moves as games launch and retire, so we are not framing any state as permanently generous or permanently stingy. But the size of the gap is durable: state portfolios differ in price mix, payout design, and how quickly depleted games leave shelves. A handful of lotteries publish less granular data than the rest, and for those our methodology page documents how the numbers are modeled from published odds.

Finding 5: the Selbee pattern still exists, in miniature

On the day of this snapshot, 25 games, slightly under 1 percent of the active set, showed more prize value remaining than the cost of the tickets left to sell. This is the late-life pattern Jerry Selbee found: most tickets have sold, the big prizes haven’t hit, and for a while the arithmetic tilts.

Before anyone quits their job: a positive expected value is an average across an enormous number of tickets, not a prediction about the next one you touch, and the condition evaporates as prizes get claimed. Some of these 25 are also small games in small states where one claimed prize flips the sign overnight. The honest takeaway is narrower and more interesting: the public data is rich enough to watch this pattern appear and disappear in near real time, across the whole country. That is a thing you can only see if something reads all 42 prize tables every day.

Where the data comes from

Scratch IQ collects remaining-prize counts from the public websites of 41 state lotteries plus Washington, D.C., once per day, and recomputes each game’s expected value from the published tier counts and overall odds. The formula, the update schedule, and the per-state caveats are documented on the methodology page. The statistics in this article are a July 12, 2026 snapshot of that dataset: 2,706 active games, 2.38 million prize-tier records, collected daily since March 17, 2026. Journalists and researchers are welcome to cite these figures with the date attached; if you need a cut of the data we haven’t published, get in touch.