Data, records and curiosities
Powerball officials suspected fraud when 110 tickets matched five numbers in a single draw. The explanation was a Long Island noodle factory — and a lesson in why ticket outcomes are not independent.
On 30 March 2005 the Powerball drawing produced 22, 28, 32, 33, 39 with a Powerball of 42. The jackpot went unclaimed. What set off alarms at lottery headquarters was the tier below it.
110 tickets matched all five main numbers.
A typical draw produced a handful. Officials in several states began checking for fraud, because the alternative explanation — that a hundred and ten people had independently arrived at the same five numbers — was not credible.
It was not credible. The tickets were not independent.
The numbers came from fortune cookies.
Wonton Food Inc., a manufacturer in Long Island City, New York, printed suggested lottery numbers on the reverse of its fortune slips. One of the number sets it printed — and printed onto thousands of fortunes distributed to Chinese restaurants across the country — was 22, 28, 32, 33, 39, 40.
Five of those six matched the main numbers exactly. The sixth, 40, was intended as the Powerball. The actual Powerball was 42.
Two digits away from a hundred and ten shared jackpots.
Powerball's second-tier prizes were fixed, not pari-mutuel, so the winners did not split a pool — each was paid in full. 89 winners received $100,000 each; 21 who had bought the Power Play multiplier received $500,000 each. The total was $19.4 million in one prize tier, in one draw (Wikipedia's account of the drawing).
After investigation, lottery officials concluded there had been no fraud and paid every claim.
The suspicion was mathematically reasonable. Powerball at the time used a 5 from 53 main matrix with a 1 from 42 Powerball. So:
Now invert it. If 110 tickets won that prize and every ticket were an independent random selection, the implied number of tickets sold is:
110 × 2,939,678 ≈ 323 million
Nothing close to that many tickets were sold for a mid-size Powerball draw in 2005. The arithmetic proves that the winning tickets cannot have been independent — which is precisely what officials concluded before they knew why. The mathematics correctly identified that something unusual had happened; it simply could not tell them the cause was a noodle factory.
Put a number on it. Take a plausible sales figure for a modest 2005 draw — say 20 million tickets — and assume every one is an independent random line. The count of second-tier winners is then Poisson with mean:
λ = 20,000,000 / 2,939,678 = 6.8
Six or seven winners expected. The probability of seeing 110 or more from a Poisson with λ = 6.8 is, using Stirling's approximation for the leading term:
ln P(X = 110) = −λ + 110 ln λ − ln(110!) = −6.8 + 210.9 − 410.3 = −206.3
which gives P ≈ 10⁻⁹⁰.
One in a thousand billion billion billion billion billion billion billion billion billion billion. Not "unlikely" — impossible, in the only sense that word is useful. And yet it happened, and no fraud occurred. That is what a broken modelling assumption looks like: the model was not slightly off, it was answering a question about a world that did not exist. The tickets were never independent, so no probability computed on the assumption that they were meant anything at all.
The practical rule generalises well past lotteries: when a calculation returns 10⁻⁹⁰ for something that visibly happened, the fault is in the assumptions, not the arithmetic.
This is the cleanest real-world demonstration of a distinction that costs players money constantly.
Your probability of winning is fixed by the game's matrix. Fortune-cookie numbers are exactly as likely as any other combination — see how lottery odds are calculated.
Your expected payout is not fixed. It depends on how many other people chose what you chose. In a pari-mutuel tier — which includes almost every jackpot in the world — 110 winners means each receives 1/110th. The prize pool arithmetic is unforgiving.
Powerball's second tier happened to be a fixed prize, so the players were paid in full and the operator absorbed a $19.4m liability instead. Had those 110 tickets matched the Powerball too, they would have divided a single jackpot 110 ways, and each "winner" would have received a small fraction of a life-changing sum.
The fortune cookie is a vivid case of a completely general problem. Ticket choices cluster, hard, around:
The defence is simple and it is the only number-selection advice on this site that has a mathematical basis: choose numbers other people do not choose. Above 31 if possible, no visual pattern, no published set. It does not raise your chance of winning by a hair. It raises what you keep if you do.
Quantify it with the number sharing risk tool, and see quick picks vs chosen numbers for why a machine-generated line is usually less correlated than a human one.
Wonton Food's numbers were printed on fortunes shipped nationwide. If the company had put 42 on those slips instead of 40, 110 tickets would have shared a jackpot — and the story would have become a case study in prize dilution rather than a curiosity.
The line between "everybody won" and "everybody won almost nothing" was one ball.
Last verified: 2026-08-29