Published · · Last reviewed · · 6 min read
English · 한국어
"One more to get even": sunk costs and what lotto really returns
You spend a month buying lottery tickets and lose KRW 100,000. "One more round and I'll at least break even." Many people feel this, and the feeling has a name: judgement tilting towards recovering money already spent (a sunk cost). This page shows with numbers why that arithmetic does not hold, and separates what two old psychology papers on losses say about the feeling from what they do not. It neither recommends nor discourages buying; it supplies the numbers needed for setting a limit.
How much more must you spend to get even?
Korean Lotto pays out 50% of sales as prizes (before tax, a statutory floor, separate from rollovers). On average a KRW 1,000 game returns KRW 500. If you are already KRW 100,000 down and spend s more, the average outcome is −100,000 − 0.5 × s won. Spending more does not shrink the average loss; it grows it.
| Additional spend | Average amount returned | Average cumulative loss |
|---|---|---|
| KRW 0 | KRW 0 | −KRW 100,000 |
| KRW 100,000 | KRW 50,000 | −KRW 150,000 |
| KRW 200,000 | KRW 100,000 | −KRW 200,000 |
| KRW 1,000,000 | KRW 500,000 | −KRW 600,000 |
There is a further trap in "average". An average of 50% does not mean you usually get 50% back.
What you usually get back
The lottery's average return is propped up by very rare large prizes. To see the usual outcome (the median) you need a simulation. I used the median prize per winner from round 88 on (first about KRW 2.03 billion, second about KRW 56.13 million, third about KRW 1.47 million, fourth KRW 50,000, fifth KRW 5,000) and repeated the purchase of n games many times (fixed-seed simulation, 100,000 repetitions for 10 and 100 games, 20,000 for 1,000 games).
| Games bought (money spent) | Median returned | 10th to 90th percentile | Mean (for reference) | Chance of at least breaking even |
|---|---|---|---|---|
| 10 (KRW 10,000) | KRW 0 | KRW 0 – 5,000 | about KRW 2,280 | 3.3% |
| 100 (KRW 100,000) | KRW 10,000 | KRW 5,000 – 55,000 | about KRW 48,700 | 1.2% |
| 1,000 (KRW 1,000,000) | KRW 175,000 | KRW 105,000 – 275,000 | about KRW 558,000 | 2.9% |
The usual outcome for someone who buys 100 games (KRW 100,000) is KRW 10,000 back. The mean of KRW 48,700 is pulled up by large prizes that turn up in only a handful of the 100,000 experiments, a first prize, or the rarer second and third. Even at 1,000 games the median is KRW 175,000, 17.5% of the money spent, and only 2.9% of cases reach break-even, mostly cases where a single third prize (about KRW 1.47 million) was hit. An average payout of 50% should be read not as "I get half back" but as "a few people get a lot and most get much less". Simulated figures vary a little with the seed and the number of repetitions, so read them as approximations.
The chance that "one more" brings a first prize
| Games bought | Money spent | Chance of at least one first prize |
|---|---|---|
| 100 | KRW 100,000 | 0.0012% |
| 1,000 | KRW 1 million | 0.012% |
| 10,000 | KRW 10 million | 0.12% |
| 100,000 | KRW 100 million | 1.2% |
| 1,000,000 | KRW 1 billion | 11.6% |
Even KRW 100 million of tickets gives a 1.2% chance of a first prize. The chance of winning some prize at least once is 21.4% in 10 games, 70.1% in 50 and 91.0% in 100. That is why the experience of "winning fairly often" builds up, but nearly all of those wins are fifth prizes (KRW 5,000), which do not cover the money spent. Use the tool below to change the probability and the number of tries and compute the chance of at least one hit.
What a year looks like
Take someone who buys 5 games (KRW 5,000) every week for a year (52 weeks, 260 games, KRW 260,000 spent). With a 50% average payout the expected return is KRW 130,000, but its make-up is this:
| Event | Chance of happening at least once in the year |
|---|---|
| Any prize | 99.8% |
| Fourth prize or better (KRW 50,000 or more) | 30.4% |
| Third prize or better (about KRW 1.47 million or more) | 0.75% |
| Second prize or better | 0.022% |
| First prize | 0.0032% |
Fifth prizes come 5.8 times a year on average (about KRW 29,000) and fourth prizes 0.35 times (about KRW 18,000). The "regular" return from the two together is about KRW 47,000, 18% of the money spent. The remaining roughly KRW 83,000 of the KRW 130,000 expected comes from the rare prizes of third rank and above. The chance of receiving none of those in the whole year is 99.25%, so this person's realistic year is "five or six KRW 5,000 prizes, a KRW 50,000 prize that may or may not come, and about KRW 210,000 lost out of KRW 260,000". These numbers are not an answer to how much is reasonable to spend a year; they are the realistic picture worth knowing when deciding.
The sunk-cost effect
Arkes and Blumer (1985) described the tendency to continue an endeavour because of resources already invested as the sunk-cost effect. According to the summary of the abstract, theatre subscribers who had paid more for a season ticket attended more plays over the following months, and in a questionnaire people who had already invested in a project rated its chance of success higher. The authors read the motive as a wish to avoid appearing wasteful. I could not open the body of the paper and checked only the abstract summary.
The point is that money already spent should not affect the next choice. The expected value of the next game is KRW 500 per KRW 1,000 whatever the number of earlier losses. "Wanting to break even" does not change that figure.
Taking risk in the face of loss
Prospect theory (Kahneman and Tversky, 1979) proposed that people feel gains and losses by different standards. According to the summary of the abstract the value function is generally concave for gains and generally convex for losses, and steeper for losses than for gains. Convexity on the loss side is read as meaning that someone already in loss may find a gamble that could make the loss bigger but could also recover it more attractive than a sure loss. Whether the wish to buy more lottery tickets fits this prediction was not tested here. I could not open the body of the paper or the size of the loss-aversion coefficient, so no figure is quoted.
Practical ways to set a limit
Rather than trying to overcome with willpower the pull of money already spent, it is more realistic to set rules before spending. Fix a monthly budget in advance and keep it separate; do not buy on a day when the reason is "to win back what I lost"; fix how much and how often you buy; keep a record so you can see the total spent. If the amount lost or the urge to buy starts to weigh on daily life, see the guidance and help organisations on the Responsible play page. This site does not sell lottery tickets and does not encourage buying.
Questions that come up
"After so many losses, isn't a win due?" No. Every game is independent, so the chance of the next game is the same however many have been lost (The gambler's fallacy).
"But fifth prizes come quite often." A fifth prize is KRW 5,000 and a game costs KRW 1,000, so fifth prizes alone return only about KRW 112 per game on average (KRW 5,000 × 182,780 ÷ 8,145,060). The rest of the return is filled by the rare first, second and third prizes.
What the simulation and the papers cannot say
The 50% payout is the statutory floor on prizes as a share of sales (before tax); tax and rollovers were not separately calculated. The simulation uses fixed tier probabilities and the median prize per winner from round 88 on, whereas real prizes depend on the number of winners. The two psychology papers were checked through abstract summaries only, and this page does not claim that those studies apply as they stand to buyers of Korean Lotto.
LottoFlick is independent and not an official lottery operator. Numbers are random and cannot be predicted.
Sources and verification
- Arkes & Blumer (1985), "The psychology of sunk cost", Organizational Behavior and Human Decision Processes 35: 124–140 (abstract summary read; full text not opened) — partly verified
- Kahneman & Tversky (1979), "Prospect Theory: An Analysis of Decision under Risk", Econometrica 47(2): 263–292 (publisher record and abstract summary; full text not opened) — partly verified
- Donghaeng Lottery: Lotto 6/45 round results, rounds 1–1247 (collected by hand 2026-10-24) — verified
- LottoFlick calculation: seeded simulation of n games, exact chance of at least one first prize or any prize — computed by LottoFlick
- LottoFlick calculation: median prize per winner by tier, rounds 88–1247 — computed by LottoFlick