Published · 2026-10-25 · Last reviewed · 2026-10-25 · 7 min read

English · 한국어

How many Lotto 6/45 jackpot winners should each round have?

Some weeks the Korean Lotto 6/45 jackpot is worth KRW 2 billion; other weeks it is KRW 420 million. The odds did not change. What changed is how many people shared the first prize. Any one ticket has a 1 in 8,145,060 chance of matching all six numbers, whichever numbers it carries. Yet two rounds with similar sales can produce 3 jackpot winners or 63. This page checks, using 1,160 official rounds, whether that scatter is what chance alone would produce.

Two findings. First, the average matches the sales-based expectation almost exactly (10,387 first-prize winners observed, 10,394 expected). Second, the round-to-round scatter is far wider than pure chance allows. The rest of the page explains what that does and does not mean. It is not advice on choosing numbers.

How the expected number of winners is computed

A Lotto 6/45 game costs KRW 1,000, so a round's total sales divided by 1,000 gives the number of games sold. Round 1128 sold about KRW 111.4 billion, or 111,364,194 games. Each game wins the jackpot with probability 1/8,145,060, so the expected number of first-prize winners is 111,364,194 ÷ 8,145,060 ≈ 13.7. Recent rounds expect about 15.

The calculation assumes that games are independent and that every combination is chosen with equal probability. If the second assumption fails, the average stays the same but the distribution changes. That difference is the subject of this page.

First we checked that the conversion itself is sound. Multiplying the games sold by the number of ways to win each prize (1, 6, 228, 11,115 and 182,780) and summing over the whole period gives:

Every ratio lies between 0.995 and 1.003. The sales-to-games conversion works, and the draw results are consistent with expectation. So the remaining question is not whether the draw is odd but whether people choose numbers evenly.

Why the analysis starts at round 88

The 1,160 rounds run from round 88 (drawn 7 August 2004) to round 1247 (24 October 2026). Rounds 1–87 are excluded because the price per game was different, so sales cannot be converted to a game count by the same rule.

That price change is documented two ways. A Kyunghyang Shinmun report of 21 July 2004 states that from round 88 (drawn 7 August) the price of one game falls from KRW 2,000 to KRW 1,000. We could not open an original notice from Donghaeng Lottery or the Ministry of Economy and Finance, so this item is marked press report. The data agree: for the 4th and 5th prizes, whose counts are large and stable, observed ÷ expected is 0.507 for rounds 1–87 and 0.998 from round 88 on (0.492 for rounds 80–87, 0.951 for 88–95). A KRW 2,000 game means half as many games, so a KRW 1,000 conversion doubles the expectation and the ratio falls to about 0.5, which is exactly what appears. The 5th-prize amount in the data also changes from KRW 10,000 to KRW 5,000 at round 88.

One data-quality note. In 486 of these rounds a sum field disagrees with count × amount. We did not use those fields and did not drop any round for this reason. Only the winner counts and total sales enter the analysis.

How pure chance would scatter

If games were independent and every combination equally likely to be chosen, the number of first-prize winners would follow a Poisson distribution, whose variance equals its mean. With 14 expected winners the standard deviation is √14 ≈ 3.7, so most rounds would land between roughly 7 and 21. Getting 63 when 13.67 are expected has a probability of about 10^-22.

To compare all rounds at once we use standardised residuals: (observed − expected) ÷ √expected. Under pure Poisson their variance is about 1. Across the 1,160 rounds it is 1.73. The mean is 0.005, essentially zero, but the spread is 73% wider than chance predicts.

The chart groups rounds by expected winners. Dots are mean observed counts, vertical lines the 10th–90th percentile range, and the shaded band is what Poisson would give (±2 standard deviations).

The dots follow the diagonal, so the average is right. But the vertical lines extend well past the band. In the 12–14 expected bin the observed standard deviation is 6.5 against 3.6 for Poisson. Round 1128 sits in that bin, but removing it does not remove the heavy right tail below.

The shape: a heavy right tail

Bars are observed rounds; the orange marks show how many rounds pure Poisson would put in each bin. The middle agrees, the ends do not. Standardised residuals of 2 or more should occur in about 36 rounds and occur in 67. At 3 or more: 21 rounds instead of 4.5. At 4 or more: 9 rounds instead of 0.4. The left end is slightly heavy too (21 rounds below −2 instead of 13), so rounds with unusually few winners are also more common than chance suggests. This is not the work of one or two outliers.

Here is the usual explanation. If everyone picked numbers independently and uniformly, the number of players on any one combination would be Poisson. But some combinations are chosen more often than others. When the drawn combination happens to be popular, many winners appear at once; when it is unpopular, few do. The average is the same, the variation between rounds is larger. Statisticians call this overdispersion.

A few unusual rounds

The most visible common feature is a high share of manual picks: 52 of 63 (83%) in round 1128, 42 of 50 (84%) in round 1019, 27 of 30 (90%) in round 546. Across all rounds with a breakdown, about 31% of first-prize winners were manual. The unusual rounds were won disproportionately by people who wrote their own numbers. The next article takes this up.

Number shapes also stand out. Round 1162 (20–29, close together) and round 1152 (30–37) are numbers that sit near each other on the play slip. Round 1128 has five numbers at or below 31. The explanation often offered, that people play birthdays, fits some rounds, but round 1152 had mostly numbers above 31 and still drew many winners. No single reason covers every case, and this page does not claim a cause. What the data support is indirect evidence that people choose numbers unevenly.

The opposite end exists too: three rounds since round 88 had no first-prize winner (rounds 289, 295 and 463). They expected about 4.8 to 6.0 winners, and the chance of zero at those levels is roughly 0.2% to 0.8%. Among 1,160 rounds a few such cases are unremarkable.

Does it change by year?

The yearly variance of standardised residuals ranges from 0.97 to 4.7: 2005 is close to Poisson, 2022 is 3.79 and 2024 is 4.69. With only 52 rounds a year, a single round such as 1128 can lift a year's variance a lot, so differences between years should not be read as a trend. Expected winners per round more than doubled, from 6.4 in 2005 to 15.0 in 2026, yet overdispersion did not move in any consistent direction. There is no sign that larger sales alone caused the wider scatter.

What this analysis does not say

• Winning probability does not change. In the round with 63 winners, each game still had a 1 in 8,145,060 chance. What differed was how many people held the same combination.
• It is not evidence that some numbers come up more often. It looks at how many people chose a drawn combination, not at which combinations are drawn. The draw results match expectation across all prize levels.
• Unusual rounds cannot be known in advance. Nobody can tell before a draw whether a popular combination will be drawn. The table explains what already happened and predicts nothing.
• The sales conversion has some error. For the 4th and 5th prizes, round-level deviation is about 4–7% of expectation, a mix of conversion error and selection skew that we could not separate. Even a 5% conversion error is 0.45 winners in a round that expects 9, which cannot explain the first-prize variance of 1.73.

See it yourself

Open the Lottery Lab link below this article and run a Lotto 6/45 simulation. In a simulation every combination has the same chance of being chosen, so first-prize counts scatter like a Poisson variable. The much wider scatter in the real rounds therefore comes from human choices, not from the draw.

The next article, Same odds, different split risk, uses the same data to look at auto, manual and semi-auto winners and the limits of reading those shares.

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

Sources and verification

• Donghaeng Lottery: Lotto 6/45 round results (public JSON, collected by hand on 2026-10-24) — verified
• LottoFlick aggregate output used on this page — computed by LottoFlick
• Kyunghyang Shinmun, 2004-07-21: price falls from KRW 2,000 to KRW 1,000 from round 88 — press report
• Dailian report on round 1128 (63 first-prize winners) — press report

Related reading

• Same Odds, Different Split Risk: Auto vs Manual Lotto 6/45 Winners The math of lotteries
• South Korea lottery prize tax guide and net payout calculator
• Combination Calculator
• Lotto 6/45 game guide