Seven million tickets
For the draw on Saturday 15 February 1992, the Virginia Lottery’s Lotto jackpot stood at more than $27 million. A play was six numbers from 1 to 44, and there are exactly 7,059,052 ways to choose them, so at $1 a play a little over $7 million would buy every possible ticket, including the one certain to match. The Lottery’s director, Ken Thorson, said on television soon afterwards that spending more than $7 million to buy every combination seemed preposterous. Somebody had just tried.
The buyer was the International Lotto Fund, an Australian trust that, in the Fourth Circuit’s words, “systematically purchased a large number of tickets” that February. It had about 2,500 investors, according to a 2019 Washington Post retrospective, and was organised by Stefan Mandel, who by his own account had won this way before: once in Romania, with a block of tickets bought with money pooled among friends, then 12 or 13 times in Australia, depending on the telling, until the rules there changed to stop him. In a 1994 interview he said the only sure way to win was to cover every possibility, and that not every lottery was worth doing.
The method was paper. By Mandel’s account to NPR, printers ran for more than a month producing pre-filled play slips, which was legal at the time, and stores agreed in advance to take stacks of 10,000 slips at once; in a 1992 television interview cited by the Post he spoke of 30 computers and 12 laser printers. The Age counts 35 people assigned to the buying.
The problem was time. Even spread over three full days and nights, 7,059,052 plays is 27.2 tickets a second (calculation for this article). Time put the spending at about $5 million and the Post put the tickets at five to six million, after one store stopped selling hours before the deadline. Call it 5 to 5.6 million tickets, 71 to 79% of the combinations; about 1.5 to 2 million were never bought.
It was enough. The six numbers drawn on 15 February were on one of the Fund’s tickets and on nobody else’s: Time reported a single winning ticket. The trustee claimed it on 5 March. The prize, as the federal courts later set it out, was a jackpot of $27,036,142, payable in 20 annual instalments, plus “numerous additional prizes for lesser amounts, totalling $662,868”.
| ITEM | FIGURE |
|---|---|
| combinations | 7,059,052 |
| every ticket | $7,059,052 |
| bought | 5–5.6M |
| jackpot | $27,036,142 |
| smaller wins | $662,868 |
| on paper | 3.83× |
Combinations: C(44, 6), at $1 a play. Jackpot and smaller prizes: the district court. Tickets bought: Time, The Age and The Washington Post, which differ. On paper: the jackpot divided by the cost of every combination, a calculation for this article.
What a full cover buys
Buying every combination is called a full cover, and it takes the luck out of the numbers. Whatever six balls are drawn, the 7,059,052 tickets hold exactly one jackpot winner and a fixed count of every smaller match.
| MATCH | TICKETS |
|---|---|
| 6 of 6 | 1 |
| 5 of 6 | 228 |
| 4 of 6 | 10,545 |
| 3 of 6 | 168,720 |
| 2 or fewer | 6,879,558 |
Calculation for this article: C(6, k) × C(38, 6 − k) tickets match exactly k of the six numbers drawn. Virginia’s 1992 prize amounts for the smaller matches are not in the record used here.
What it cannot take out is everything that happens after the draw. Virginia’s jackpot was paid over twenty years, it was subject to tax withholding, and it would have been split equally with anyone else holding the winning numbers. Each of the three takes a piece of the 3.83.
What $27 million was worth
Start with time. The jackpot came as 20 annual instalments of $1,351,807, the court’s figure divided by twenty; at an 8% discount rate they were worth about $14.33 million at the draw, 53 cents on the headline dollar. Then tax: the Lottery said it would withhold 30% of each payment for federal tax and 4% for Virginia, leaving $892,193 a year. Whether any of that was later refunded is not in the record, so the figures here assume it was withheld.
| RATE | VALUE | AFTER 34% | VS COST |
|---|---|---|---|
| 6% | $16.44M | $10.85M | 1.54× |
| 7% | $15.32M | $10.11M | 1.43× |
| 8% | $14.33M | $9.46M | 1.34× |
| 9% | $13.45M | $8.88M | 1.26× |
Calculation for this article: twenty instalments of $1,351,807, the first paid at once, discounted at each rate; after 34% removes the withholding; vs cost divides that by $7,059,052. Overheads are ignored.
So the 3.83 on paper is about 2.03 once the payments are valued at 8%, and about 1.34 once the withholding comes off: still a profit for a sole winner holding every ticket, but the return has shrunk from nearly four times the money to about a third more, before a cent of overheads.
The break-even rule
Run the arithmetic backwards and it gives a rule of thumb for any jackpot paid and taxed this way: buying every combination pays a sole winner only when the jackpot is at least about 2.9 times the cost of the full set. One co-winner halves the prize, so it doubles the bar, to 5.72 times at 8%.
| RATE | SOLE WIN | 1 CO-WINNER |
|---|---|---|
| 6% | 2.49× | 4.98× |
| 7% | 2.67× | 5.35× |
| 8% | 2.86× | 5.72× |
| 9% | 3.05× | 6.09× |
Calculation for this article: 1 ÷ (what a headline jackpot dollar is worth after discounting the twenty instalments and withholding 34%). Smaller prizes, which help, and overheads, which hurt, are left out.
Virginia’s 3.83 cleared the first bar and fell far short of the second: half the jackpot would have made the full cover worth 0.67 times its cost, a third of the money gone on the night the plan worked. NPR’s 2018 report has Mandel waiting for jackpots of $25 million or more so that he would profit even if he had to split one. At face value half of $25 million is 1.77 times the cost of every Virginia combination; paid over twenty years, valued at 8% and with 34% withheld, it is worth 0.62 times (calculation for this article).
The ticket nobody else held
Whether the Fund would have to share was out of its hands: anyone else playing that week could hold the winning six. How likely that was depends on how much of the field the other players held, which the court record does not say, so the table prices the full cover the Fund set out to buy against that unknown. It assumes the other tickets were spread evenly at random, which flatters the Fund: players doubling up on the same combinations would split a shared jackpot more ways.
| OTHERS HOLD | AVG SHARE | WORTH |
|---|---|---|
| none | 100% | 1.34× |
| 25% | 87% | 1.16× |
| 50% | 72% | 0.97× |
| 75% | 54% | 0.73× |
| 90% | 39% | 0.52× |
Calculation for this article. Others hold: the share of all combinations held by at least one other ticket. Avg share: the Fund’s expected share of the jackpot, co-winners counted as Poisson. Worth: the jackpot alone at 8% after 34% withholding, against the $7,059,052 cost; the smaller prizes would add about 0.08 to 0.09.
The break-even comes at around half the field: 46% counting the jackpot alone, about 56% with the smaller prizes added. The jackpot multiple says whether buying every ticket can pay. The crowd decides whether it does.
The position the Fund actually held carried a second risk. With 71 to 79% of the combinations, there was a 21 to 29% chance before the draw that the winning numbers were among those it never bought. A sole win needed the Fund to hold them and nobody else to: about a 35 to 40% chance if others held half of all combinations, 18 to 20% if they held three quarters (calculation for this article).
| TICKETS | HELD | SOLE WIN | SPLIT IN 2 |
|---|---|---|---|
| 5.0M | 71% | 1.98× | 1.04× |
| 5.5M | 78% | 1.80× | 0.94× |
| 5.6M | 79% | 1.77× | 0.93× |
Calculation for this article. Held: the share of the 7,059,052 combinations bought. Sole win and split in 2: the jackpot at $14.33 million (8%) after 34% withholding, plus the $662,868 of smaller prizes less an assumed 30%, as multiples of the ticket cost. Overheads are ignored.
What happened, the only winning ticket, was worth about 1.8 to 2 times what the tickets cost, after discounting and withholding. One other winning ticket anywhere in Virginia would have left the Fund near break-even before overheads. Buying most of the combinations had turned a one-in-7,059,052 ticket into roughly a three-in-four chance of holding the jackpot numbers. Whether holding them paid was still a lottery.
Pay the tax, then sue
The money came with a fight. On 10 March 1992 the Lottery said it would pay, withholding 30% for federal tax on the recommendation of the Internal Revenue Service and 4% for Virginia tax; the Post says it also briefly tried to stop the payout altogether. The Fund answered with a tax treaty. Under Article 7(1) of the 1982 U.S.–Australia income tax treaty, an Australian enterprise’s business profits are taxable only in Australia unless it has a permanent establishment in the United States. Playing lotteries at scale was the Fund’s business, it argued, so nothing should be withheld.
- 14 August 1992: the Lottery’s director formally refused the exemption from both federal and state withholding.
- 28 August 1992: District Judge Richard L. Williams, in Richmond, ordered the Lottery to accept the Fund’s exemption form and not to withhold Virginia tax; a permanent injunction followed, requiring every instalment to be paid in full.
- 31 March 1994: the Fourth Circuit, on an appeal by the Lottery joined by the United States, reversed and sent the case back “with directions to dismiss it for lack of jurisdiction”.
The appeals court never decided whether the treaty applied. In an opinion by Judge Wilkinson it held that withholding is tax collection, so the Anti-Injunction Act barred an injunction against the federal withholding and the Tax Injunction Act barred one against Virginia’s, since the state’s courts offered a refund remedy. The Fund’s route was “to pay the tax under protest and then seek a refund”. Courts still cite the case on tax injunctions, the Ninth Circuit among them in 2016; no later published opinion involving the Fund turns up, so how any refund claim ended is unknown.
Investigators looked as well. The Post reports that the Australian Securities Commission and the FBI examined the operation, and The Age’s 2011 column says an inquiry involving American and Australian law enforcement proved no wrongdoing; no charges were reported. Buying in bulk with pre-filled slips broke no rule at the time. What officials were considering, Time reported, was a rewrite of the rules to stop well-funded players cornering a draw and driving off everyone else. Virginia’s current regulations require play slips to be marked by hand, or electronically where a game allows, never by an “electro-mechanical, electronic printing, or other automated device”, so the Fund’s computer-printed slips would be invalid today; when that rule first appeared, the sources for this article do not say.
A full cover at a fixed-odds table
Strip the story down and a full cover is the average ticket made certain: buy every ticket at once and the luck cancels, leaving only the prices. So it pays only when the average ticket is worth more than it costs, which in a lotto takes a prize pool larger than the cost of every combination and few others to share it. Cash WinFall’s roll-down was the same kind of surplus, released by a rule instead of a jackpot.
A fixed-odds game has no pool to grow. Keno makes the cleanest comparison, because it is a lottery draw with the prices printed: forty numbers, ten drawn, a multiplier for every hit count.
Price a full cover of it the same way. Forty $1 one-pick tickets cover every number; on any draw exactly ten are hit and pay $3.96 each, so $40 returns $39.60, every time, and all 780 two-pick tickets return $772.50 on $780 (calculation for this article). Every table works the same way: the 98.65% to 99.07% returns of the forty keno tables (keno, priced) are exactly what a full cover gets back. At a fixed-odds table, buying every ticket is not a strategy. It is the house edge with the luck taken out.
Each keno bet gets its own draw, so that position cannot even be placed; forty separate one-pick bets merely average the same $39.60. What a keno table can never do is what Virginia’s jackpot did in 1992: grow past the price of every ticket. Nothing carries over between rounds, nobody shares your prize, and no amount of buying changes the sign; how many bets before the edge shows prices how long that sign takes to appear.
Mandel’s line that the only sure way to win is to cover every possibility was true of the numbers and not of the money. In Virginia a full cover was exposed to three things no ticket could buy off: a twenty-year wait, the tax, and anyone else holding the same six numbers. For a sole winner with every ticket the edge was about a third, and the Fund came within one other winning ticket of losing it.
A lottery draw with the prices printedForty numbers, ten drawn, the multiplier for every hit count on screen before you betKeno →
Who is Stefan Mandel?
A Romanian-born organiser of lottery syndicates who, by his own account, won once in Romania with a block of tickets and then 12 or 13 times in Australia by buying every combination, until the rules there changed. In February 1992 the International Lotto Fund he organised, an Australian trust with about 2,500 investors, set out to buy every combination of the Virginia Lotto and won its $27 million jackpot.
Did Stefan Mandel buy every ticket in the Virginia lottery?
No. The Fund set out to buy all 7,059,052 combinations but ran out of time after about 5 to 5.6 million tickets, 71–79% of them; sources differ on the exact number. Its tickets still included the only jackpot winner of the 15 February 1992 draw: $27,036,142 in 20 annual instalments, plus $662,868 in smaller prizes.
When does buying every lottery ticket pay?
Only when the jackpot is several times the cost of every combination and few or no other tickets share it. For a 20-year annuity valued at 8% with 34% withheld, a sole winner needs a jackpot of about 2.86 times the cost of the full set; one co-winner doubles that to 5.72. Virginia’s jackpot was 3.83 times the cost: worth about 1.34 times for a sole winner and 0.67 with one co-winner (calculation for this article).
Was buying every combination legal?
At the time, yes: pre-filled play slips were legal, and the Lottery paid. The FBI and the Australian Securities Commission examined the operation, no charges were reported, and a 2011 account says no wrongdoing was proven. Virginia’s current rules forbid marking play slips with an automated printing device.
What did the court decide about the tax?
A federal district judge in Richmond ordered the Lottery to pay each instalment in full, without its 34% withholding, but in 1994 the Fourth Circuit reversed for lack of jurisdiction: withholding is tax collection, so the Anti-Injunction Act and the Tax Injunction Act barred the order, and the Fund’s route was to pay under protest and seek a refund. No court ruled on its treaty argument.
Does buying every combination work at a casino game?
Not for profit. At a fixed-odds game a full cover returns exactly the table’s return: forty one-pick keno tickets on one draw always include exactly ten hits, $39.60 back on $40 on this site’s Classic table. Only a prize pool larger than the cost of every combination can lift a full cover above its price.
- International Lotto Fund v. Virginia State Lottery Dep’t, 800 F. Supp. 337 (E.D. Va. 1992) — the jackpot and lesser prizes, the claim, the withholding notice, the treaty claim, the injunctions
- International Lotto Fund v. Virginia State Lottery Dep’t, 20 F.3d 589 (4th Cir. 1994) — the Fund, the 20 instalments, the denial, the reversal
- Collins Holding Corp. v. Jasper County, 123 F.3d 797 (4th Cir. 1997), and the CourtListener citation index (Fredrickson v. Starbucks Corp., 9th Cir. 2016) — later citations
- Time, 9 March 1992 — the jackpot, the 7,059,052 combinations, about $5 million spent, one winning ticket, the proposed rule changes
- NPR, Planet Money on All Things Considered (Alex Goldmark), 23 October 2018 — Thorson on television, the pre-filled slips and 10,000-slip stacks, the earlier wins, the $25 million rule
- The Age, Burkard Polster and Marty Ross (Monash University), 19 September 2011, from the authors’ column archive — 35 buyers, around five million tickets, no wrongdoing proven
- The Washington Post (2019), republished by the NZ Herald — five to six million tickets, about 2,500 investors, the 1992 television interview, the inquiries, the attempt to stop the payout
- Bursa.ro (2012), republishing Mandel’s 1994 interview — covering every possibility, not every lottery worth doing
- Virginia Administrative Code, 11VAC5-41-110 — play slips may not be marked by an automated device
- Betkyo engine source: BALLS and MAX_PICKS in keno/kenoGame.tsx; kenoDraws() and demoKenoBet() in _shared/demoLocal.ts; KENO_PAYTABLE in keno/paytable.ts
- Calculation for this article: C(44, 6) and the match counts of a full cover; present values of twenty instalments at 6–9%; 34% withholding on the jackpot and an assumed 30% on smaller prizes; break-even multiples; Poisson co-winner shares; the keno full covers and the range of the forty tables



