Published · · Last reviewed · · 5 min read

English · 한국어

Is there a "lucky store"? Sales volume and jackpot counts

You may have seen a banner outside a lottery retailer announcing "first prize won n times". People call such places lucky stores and travel to them on purpose. This page does not rate or promote any retailer, so it names none. It concentrates on one question: if every store had exactly the same chance, would some stores still win the jackpot several times, and what data would be needed to tell whether a store is really special?

A win attaches to a game, not to a store

A game has a 1 in 8,145,060 chance of the jackpot wherever it was bought. A store does not create numbers and takes no part in the draw; it prints a game and hands it over. So a store that sells twice as many games is expected to have twice as many jackpots. Busy places, such as near stations or terminals or with loyal regulars, sell more. Frequent jackpots there are a product of sales volume, not luck.

That is why "first prize n times" alone settles nothing. What is needed is the number of games the store sold in the same period. It is the same problem as in "Auto wins more often": what that statistic really means: a ratio without a denominator. Store-level sales are not public, so this site does not know them.

Even identical stores produce leaders

Take numbers. Assume 5,000 stores with identical sales and 1,000 jackpots in total. (The store count and number of wins are assumptions for illustration, not real figures.) Each store is expected to win 0.2 times. With a Poisson distribution:

Assumption: 5,000 stores, identical sales, 1,000 jackpots (0.2 expected per store)
A store wins k or more timesProbability for one storeExpected number of such stores
1 or more18.1%about 906
2 or more1.75%about 88
3 or more0.115%about 5.7
4 or more0.0057%about 0.28 (chance that at least one exists is about 25%)

Although all stores are identical, about 900 win once, 88 win twice and five or six win three times. A few could hang a "three-time winner" banner. That is normal. Because the same happens when every store is the same, a banner count is not evidence that a store is special. Change the numbers in the tool below.

This tool needs JavaScript.

The luckiest store's count changes each run, but a "lucky" store with around three wins nearly always appears. To see unequal sales, imagine a store selling ten times more: expected wins rise from 0.2 to 2, the chance of three or more is 32% and of five or more 5.3%. Big sellers filling the top of the leaderboard is just what probability predicts.

Selection effect: we only see the winners

Store promotion has another trick. Thousands of stores that never paid a jackpot hang no banner, so what we see is only the winners, a survivorship bias. With the hidden denominator (the number of stores with no jackpot and their sales) missing, every comparison makes every winning store look special.

And once a store is known, more people go there. Sales rise, expected jackpots rise with them, and the belief in a lucky store produces its own evidence. That increase is an effect of volume, not of the numbers or the location being lucky.

What a proper test would need

  1. Games sold by each store over the same period
  2. Jackpots (and ideally lower prizes) won by each store
  3. A benchmark that ignores which store sold the ticket (expected wins = games sold × the probability of each prize)

Item 1 is not public. The national open-data portal does publish a list of first-prize winning stores (the Ministry of Planning and Budget's "Online lottery first-prize winning store list", updated annually, licence shown as "no restriction on use", checked 2026-10-25). We did not use it and discuss no store. The list gives names, regions and win counts but not sales, so it cannot support the test above. Adding second and lower prizes would raise statistical power, but the missing sales denominator would remain.

A calculation with two hypothetical stores

Compare two stores with different sales. (All numbers are assumptions.) Store A sells 10,000 games per round and store B 2,000. Over ten years (520 rounds), the expected number of jackpots from a store is games sold × rounds ÷ 8,145,060.

Assumption: expected jackpots over ten years (520 rounds)
StoreSales per roundExpected jackpotsChance of one or more
A10,000 games0.63847.2%
B2,000 games0.12812.0%

With five times the sales, the chance of producing a jackpot within ten years is about four times higher (47.2% against 12.0%). A ticket from A and a ticket from B are the same 1 in 8,145,060. If word spreads that A is "lucky", sales rise further, the gap widens, and the rumour produces its own evidence. The difference between the stores is not luck per ticket but the number of tickets sold.

Questions that come up

Is buying at a "lucky" store really pointless? As far as probability goes, yes: a ticket's chance is the same anywhere. Choosing a store because of preference or meaning is a personal matter, and this page does not judge that choice.

How should a high jackpot count be read? Most accurately as "a lot was sold there". Without knowing the sales volume, you cannot tell whether the count is high or low.

Does buying online remove lucky stores? The probability of one game is the same for online purchases too. A different place of sale does not change the probability.

If store data were used: an analysis design

Here is how a test should be designed, assuming sales data existed. This site has not carried it out, and it makes no claim about results.

  1. Compute each store's expected jackpots: games sold in the period ÷ 8,145,060.
  2. Take the difference between actual and expected jackpots and divide by the square root of the expected number to get a standardised value.
  3. Do the same for every store, then compare whether the number of stores with extreme values exceeds what chance would give.

Step 3 matters. With 5,000 stores and no real difference at all, about 5,000 × 0.0027 ≈ 13 stores will exceed a standardised value of 3. This is the multiple-comparisons problem. One should not pick the few most striking stores and say "this cannot be explained by chance"; the question is whether the distribution over all stores differs from the chance distribution. Using only first prizes also leaves too few events, so including the second to fifth prizes is statistically better. All of this requires the denominator of store-level sales.

Limits and summary

In short, "first prize n times" shows sales volume and chance, not a store's character. That a large count does not raise the chance of the next ticket is the same reasoning as in the gambler's fallacy.

LottoFlick is independent and not an official lottery operator. Numbers are random and cannot be predicted.

Sources and verification