Scratch cards guide
Cosmos Scratch: finite runs and prizes already claimed
Cosmos is only the tag on this page; it refers to no space-themed card. The page deals with a feature that sets paper scratch cards apart from most games of chance: the print run is finite, so every ticket sold changes what is left.
Drawing without replacement
A die or a coin has no memory. A print run does, in a narrow sense: once a winning ticket is sold, it is no longer in the pool. Statisticians call this sampling without replacement, and the right tool for it is the hypergeometric distribution. It answers questions of the form: if a pool of N items contains K special ones, what is the chance of finding exactly k of them in a sample of n?
The example uses a hypothetical run of 300,000 tickets at 2 units each, with four top prizes of 10,000 units.
| Prize | Tickets | Odds | Expected value per ticket |
|---|---|---|---|
| 10,000.00 | 4 | 1 in 75,000.0 | 0.1333 |
| 1,000.00 | 20 | 1 in 15,000.0 | 0.0667 |
| 100.00 | 300 | 1 in 1,000.0 | 0.1000 |
| 20.00 | 2,500 | 1 in 120.0 | 0.1667 |
| 5.00 | 15,000 | 1 in 20.0 | 0.2500 |
| 2.00 | 48,000 | 1 in 6.2 | 0.3200 |
65,824 winning tickets out of 300,000 (overall 1 in 4.56). Prize fund 311,000.00 on sales of 600,000.00: return 51.83%, so a ticket at 2.00 is worth 1.0367 on average. Hypothetical print run.
The top prizes in a purchase of 20
Buy 20 tickets from a fresh, unsold run. The chance that none of the four top prizes is among them is hypergeometric with N = 300,000, K = 4, n = 20, k = 0. That gives a chance of at least one top prize of 0.026664%. The independent-draw shortcut gives 0.026663%, which shows how little the difference matters for small purchases from a large run.
Half the run is sold
Now suppose 150,000 tickets have been sold. Before anyone checks, how many top prizes are likely to have gone with them? The hypergeometric distribution answers that directly.
| Top prizes already sold | Chance |
|---|---|
| 0 | 6.25% |
| 1 | 25.00% |
| 2 | 37.50% |
| 3 | 25.00% |
| 4 | 6.25% |
Assumes tickets are sold in random order from a well-shuffled run.
There is a 31.25% chance that three or all four are already gone at the halfway point. Many issuers publish a list of prizes still unclaimed, and that list is the only way a buyer could know which case applies.
What the remaining tickets are worth
Assume the smaller tiers sell in proportion to the run, so half of them are gone too. The expected prize on one of the 150,000 unsold tickets then depends on how many top prizes remain.
| Top prizes left | Expected prize per ticket | Return on a 2-unit ticket |
|---|---|---|
| 4 | 1.1700 | 58.50% |
| 2 | 1.0367 | 51.83% |
| 0 | 0.9033 | 45.17% |
At the start the expected prize is 1.0367 per ticket. All rows assume the lower tiers are half sold.
The spread is real but modest. Even with every top prize still waiting, the unsold tickets return well under their price. With none left, the return drops further, while the card in the shop looks exactly the same.
Limits of the idea
- A buyer cannot see which tickets in a pack are winners, so the finite run gives no edge inside a single pack.
- Unclaimed prize lists can lag behind sales, and a sold ticket is not always claimed at once.
- The effect matters most late in a run. Early on, the run behaves almost like independent draws.