Published · · Last reviewed · · 7 min read
English · 한국어
Same odds, different split risk
"Most jackpot winners play auto" is a common line. It leads some people to conclude that auto tickets win more, and others to argue that hand-picked numbers are better. Both misread the same figure. This page splits 9,336 first-prize winners across 983 Lotto 6/45 rounds into auto, manual and semi-auto, then separates what those shares show from what they cannot show. It does not recommend how to pick numbers.
The previous article, How many Lotto 6/45 jackpot winners should each round have?, showed that first-prize counts scatter much more than chance alone allows. Here we look at how auto and manual picks relate to that scatter.
What was counted
The range is round 262 to round 1247, restricted to rounds with a first-prize winner whose auto, manual and semi-auto counts add up to the first-prize total: 983 rounds. Before round 262 the breakdown is not in the data. We matched the fields to labels using round 1128, where a press report gives "11 auto, 52 manual" and the data fields read 11 and 52. That mapping rests on one report, so it is marked press report; we could not open Donghaeng Lottery's field documentation.
| Selection type | Winners | Share |
|---|---|---|
| Auto | 6,187 | 66.3% |
| Manual | 2,885 | 30.9% |
| Semi-auto | 264 | 2.8% |
Auto takes two thirds, which is the source of the line above. What matters is what this number is compared to.
A share without a denominator
Auto being 66% of jackpot winners does not show that auto games win more often. If auto accounts for 66% of all games sold, 66% of winners is exactly what we should expect. If auto were 80% of games, 66% would even suggest manual games do better. The missing number is the share of games sold that were auto.
That figure is not verified: we found no published breakdown of auto and manual sales. So this site cannot conclude that either type is favoured, and any source that does has filled the denominator with an assumption.
There is also a more basic reason. The draw machine ignores how a ticket was filled in. An auto game and a manual game each win the jackpot with probability 1 in 8,145,060. No mechanism lets the selection method change a single game's odds. What can change is how many other people hold the same combination.
Numbers make the point. Let s be the share of all games sold that were auto, and take 66.3% of winners as auto. How many times better did an auto game do than a manual one?
| Assumed auto share of sales | Relative win rate of an auto game (manual = 1) |
|---|---|
| 50% | 1.97× |
| 66.3% | 1.00× |
| 70% | 0.84× |
| 80% | 0.49× |
From the same 66.3%, the answer runs from "auto is twice as good" to "manual is twice as good", depending on the assumption. Because the draw gives every game equal odds, one can infer that if auto were really about 66% of sales, the answer would be 1.00×. That is an inference, not published data.
Why the shares are still interesting: a clue about clustering
Shares do not reveal probability, but they reveal something else. Look at how the manual share changes with the number of winners in the round.
Show the numbers
| First-prize winners in the round | Rounds | Winners | Auto | Manual | Semi-auto |
|---|---|---|---|---|---|
| 1 | 12 | 12 | 75.0% | 25.0% | 0.0% |
| 2-3 | 59 | 162 | 80.2% | 19.1% | 0.6% |
| 4-7 | 314 | 1800 | 72.4% | 25.8% | 1.7% |
| 8-15 | 492 | 5253 | 68.0% | 29.1% | 3.0% |
| 16+ | 106 | 2109 | 55.7% | 40.7% | 3.6% |
In rounds with a single winner (12 rounds) the manual share was 25%; with 2–3 winners 19%; with 4–7 winners 26%; with 8–15 winners 29%; with 16 or more (106 rounds) it rises to 41%. The more crowded the jackpot, the larger the manual share. The unusual rounds in the previous article showed the same thing (83% in round 1128, 84% in round 1019).
The simplest explanation: people who choose numbers by hand tend to choose similar numbers, and when those numbers are drawn, many win at once. Auto picks scatter across combinations, so whichever combination is drawn, the number of auto winners stays similar. In crowded rounds the manual share therefore grows. This does not mean manual games win more. It means manual games are concentrated on certain combinations, so manual wins are rare but arrive together.
A caution: because rounds were grouped by the number of winners, crowded rounds may simply be those where a manually favoured combination was drawn. The chart shows association, not causation. The association is not weak, though: rounds with 16 or more winners are 10.8% of all rounds yet contain 22.6% of all jackpot winners.
Split risk in numbers
A jackpot pool is divided among its winners. Suppose you hold a winning ticket and the number K of other winners is Poisson with mean μ. Your share of the pool is 1/(1+K), and its expected value is (1 − e−μ) ÷ μ. Take a round where an average combination would have 14 winners. "Popularity multiple" is how many times more people chose your combination than an average one.
| Popularity multiple | Average other winners | Expected share |
|---|---|---|
| 0.5× (less popular) | 7 | 14.3% |
| 1× (average) | 14 | 7.1% |
| 2× | 28 | 3.6% |
| 4× | 56 | 1.8% |
The same first prize can be worth eight times more or less to you in expectation. Change the numbers in the tool below.
This is a simple model. It reflects only that the pool is divided by the number of winners. Tax is not included; see the South Korea lottery prize tax guide.
Two real rounds, two very different shares
In round 1128 (13 July 2024) the first-prize pool was about KRW 26.46 billion and 63 winners split it, each receiving about KRW 419.9 million before tax. Round 1247 (drawn 24 October 2026) had a similar pool of about KRW 29.10 billion, but only 8 winners, so each received about KRW 3.64 billion before tax. The pools were close, yet one winner's share was 8.7 times larger.
This is not a difference in probability. In both rounds a single game had a 1 in 8,145,060 chance. In round 1128 a combination many people had chosen was drawn; in round 1247 it was not. Tax widens the gap further: the part above KRW 300 million is taxed at a higher rate (see the tax guide), so the unequal split shows up again in take-home amounts.
Two cautions. Pools are not always similar; they depend on sales and rollovers. And these two rounds were chosen as extreme examples; an average round's share lies between the two values.
Do "birthday numbers" really overlap more?
If people play birthdays, numbers 1–31 will be chosen more often. The data allow a rough check. Group rounds by how many of the six drawn numbers are 31 or below, and divide observed first-prize winners by expected winners.
| Numbers ≤ 31 | Rounds | Ratio | Chance of this count in a random draw |
|---|---|---|---|
| 0–1 | 13 | (too few rounds) | 0.8% |
| 2 | 58 | 0.962 | 5.7% |
| 3 | 242 | 0.938 | 20.1% |
| 4 | 408 | 1.048 | 35.2% |
| 5 | 338 | 0.972 | 29.2% |
| 6 | 101 | 1.070 | 9.0% |
The 439 rounds with five or more such numbers have a ratio of 0.995, against 0.940 for the 313 rounds with three or fewer, a difference of 0.056. A permutation test (20,000 shuffles, seed 12345) gives a one-sided p of about 0.05. The direction matches the "birthday" idea, but the evidence is weak: it is one of several possible groupings (a multiple-comparisons problem) and round 1128 alone has a large influence. The six-numbers group, 101 rounds, is only 1.07. This site does not present the result as proof that birthdays cause overlap. The accurate summary is "plausible, but not established by these data".
How to read semi-auto
Semi-auto means you pick some numbers and the machine fills the rest. In the data its share is 0% through 2014 and 3–5% from 2015 onward. Either the type did not exist earlier, or it was counted under another type; we could not tell which. So this page does not interpret year-by-year changes in the semi-auto share.
Limits
- Sales shares by selection type are not public. Per-game win rates cannot be computed, so no type can be called favourable.
- The definition of "winner" is not confirmed. One person holding several winning games may be counted per game rather than per person. We used the counts as published.
- Association, not causation. Grouping rounds by winner count introduces a selection effect.
- Expected value stays negative. Reducing split risk leaves the chance of winning at 1 in 8,145,060. A larger share matters only if you win; it does not change the likelihood of winning or the lottery's expected return.
What to take away
Two things should be kept apart. Every set of numbers has the same chance of winning. Whether a winner collects alone or splits depends on how many others chose the same numbers. A computer that scatters picks creates little clustering; hand-picked numbers cluster more easily. That clustering does not raise or lower the chance of winning. It changes only the split risk. The generator on this site is the same in this respect: its numbers are random and give no better chance.
The simulator linked below runs real Lotto 6/45 probabilities. For how a prize is then taxed, continue to the tax guide.
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 (first-prize winners by selection type) — verified
- LottoFlick aggregate output used on this page — computed by LottoFlick
- Dailian report on round 1128: 11 auto and 52 manual first-prize winners (used to check the field mapping) — press report · field-to-label mapping rests on this report only