Are some lottery numbers more likely?
Be honest: have you ever picked numbers thinking "the 13 hasn't come up in ages, it's overdue"? Or the opposite, "the 6 is hot lately, I'll take that one"? You're not alone. The intuition feels logical, but it's mathematically wrong, here's why.
The cold truth first
Past draws have zero influence on the next draw. Not "barely". Not "a little". Zero. Period. The drum has no memory.
Each lottery draw is an independent random event. Whether the 13 hasn't been drawn for five years or the 6 came up in the last three weeks, on the next pull, every ball has exactly the same chance: 1 in 49 for the first slot, then 1 in 48, and so on, mathematically independent of anything before.
The simplest example: flipping a coin
Imagine flipping a fair coin. It lands heads ten times in a row. What comes on the eleventh flip?
Most people think: "It HAS to be tails now, ten heads in a row is super unlikely." That part is true, ten heads in a row is super unlikely (1 in 1,024). But that was true before the first flip. Now that the ten heads have already happened, the chance of heads on the eleventh flip is exactly 50%. And the same for tails.
The coin doesn't know what came before. It has no memory. It also doesn't know it "should produce tails now". Every single flip is an isolated 50/50 decision. Same logic applies to the lottery balls in the drum, except instead of 50/50 it's 49 balls, each with equal chance.
This thinking error has a name: the Gambler's Fallacy. It's so deeply wired into the human brain that practically everyone has fallen for it at some point. You have. We have. It's fine, now you know.
"But the stats show different frequencies?"
Correct. Here are the real numbers from our database. Since 9 October 1955, there have been 5.007 main draws of German Lotto 6aus49. With 6 numbers per draw, that's a total of 30.042 individual ball pulls. Spread across 49 numbers, the expectation per number is roughly 613.
What we actually observe:
| Position | Number | Times drawn | Deviation from expectation |
|---|---|---|---|
| 1 hottest | 6 | 674 | +9.9% |
| 2 hottest | 49 | 659 | +7.5% |
| 3 hottest | 26 | 652 | +6.3% |
| 1 coldest | 45 | 546 | -10.9% |
| 2 coldest | 13 | 558 | -9.0% |
| 3 coldest | 21 | 572 | -6.7% |
Looks like a difference, right? The hottest number was drawn 128 more times than the coldest one, about 20.9% spread. Sounds like a lot.
But it isn't. With 5.007 draws, statistics (the binomial distribution, if you like math terms) predicts a standard deviation of about ±23 draws per number purely by chance. When you look at 49 numbers and pick the extremes top and bottom, a 100–130-draw spread is exactly what the math expects, whether the drum has memory or not.
Put differently: this distribution isn't too skewed, it's actually too uniform for "random with memory". If the drum really did favor "hot" or "cold" numbers, the differences would be much more drastic. The data shows the opposite of what intuition tells us.
How fair is the lottery drum? What 70 years of data show
The standard deviation for a single number is only one side of the story. The more honest question is this: if you test the entire distribution across all 49 numbers, is it compatible with a fair drum? There is a standard statistical test for that, the chi-square goodness-of-fit test. We applied it to the German 6aus49 data since 1955. Here are the results.
The overall test
Across 5.007 draws with 6 numbers each, the chi-square test over all 49 numbers gives a value of χ² = 49.08 with 48 degrees of freedom, p-value = 0.4296. What does that mean in normal language?
A perfectly fair drum would, on average, produce exactly this kind of deviation. The p-value of 0.43 says: if the drum were perfectly fair, we would see a deviation at least as large as the observed one in 43 out of 100 cases. So the observed distribution is not just „somehow okay". It is statistically unremarkable in the classic textbook sense. There is no reason to reject the assumption of a fair drum.
What if we try anyway?
When 49 individual numbers are tested separately, chance alone will always produce a handful that look „suspicious". At the usual 5 percent significance level, you would expect 2 to 3 false positives on average across 49 tests. We observed 3 numbers more than two standard deviations away:
- 45 at -2.89σ (cold number, drawn too rarely)
- 6 at +2.63σ (hot number, drawn too often)
- 13 at -2.38σ (cold number, drawn too rarely)
Across 49 tests, you expect 2 to 3 such outliers on average, and we observed 3, exactly the order of magnitude randomness predicts. Anyone now saying „Aha, so 45 really is cold!" is making the classic multiple-testing mistake: if you run 49 tests at once, you will always find a few outliers, but that proves nothing. If you correct for this with the Bonferroni method (the threshold is divided by the number of tests), the significance threshold is no longer p = 0.05, but p = 0.00102. Not a single one of the 49 numbers survives this correction. Not 45. Not 6.
But what if the drum changed over the decades?
Fair skepticism. 70 years is a long time, and the lottery system has changed several times along the way. So we split the data into four separate eras and tested each one on its own:
- 1955 to 1979: classic 6aus49 without Zusatzzahl
- 1980 to 2012: with Zusatzzahl
- 2013 to today: with Superzahl (separate pot)
- Since 18 January 2023: after the first equipment change of the drawing machine in more than 60 years
Across all four eras, the chi-square test remains unremarkable. There is no indication that any version of the drum was ever measurably unfair. By the way: the German 6aus49 draws have always been carried out mechanically, never by random number generator. That is a deliberate design choice by the Deutscher Lotto- und Totoblock for maximum traceability.
A methodological note for the precise
Statistics purists rightly point out that the assumptions of the chi-square test are not perfectly met in lottery draws: the 6 numbers in a draw are drawn without replacement, so they are not fully independent of one another. With 6 out of 49, however, this dependency is so small that the test remains meaningful in practice. A strictly correct analysis using multinomial tests leads to the same conclusions.
The punchline
The lottery drum is not just fair. It is measurably fair in the data. Across 70 years, across 5.007 draws, across four drum eras. Statistics cannot squeeze much more out of it than that.
A concrete lottery example
Say you watch the last 50 draws. The 13 only came up once in that window, the 26 came up eight times. Is the 13 "overdue" now?
The honest answer: the drum doesn't know what the last 50 draws did. It doesn't "catch up" deliberately. On the next draw the 13 has the same chance as any other ball - roughly 1 in 8 of landing among the six main numbers. The 26 has the same chance. So does the 7. So does any other number.
The human brain is evolutionarily wired to spot patterns, even where there are none. A streak like "13 hasn't come up in a while" looks like a signal to our brain. To the physical drum it's nothing. The drum runs every draw the same, completely unimpressed by our intuition.
Which lottery numbers come up together most often?
A classic question in every lottery forum. The honest answer first: no combination has a statistical advantage either. But the data is fun. Here are the five number pairs that have appeared together most often in a single draw since 1955:
| Rank | Pair | Times drawn together |
|---|---|---|
| 1. | 6 and 33 | 93 times |
| 2. | 32 and 43 | 91 times |
| 3. | 10 and 11 | 88 times |
| 4. | 24 and 31 | 88 times |
| 5. | 20 and 33 | 88 times |
Expected per pair: roughly 64 times in 5.007 draws. The top pair lands at 93 times, about 45% above average. Sounds dramatic, but with 1,176 possible pairs that's exactly what the math predicts for the extreme value. In other words: 0% probability this pair shows up together in the next draw just because it did so often in the past. The drum has no memory for pairs either.
Want it one level juicier? Here are the five three-number combinations that came up together most often since 1955:
| Rank | Triple | Times drawn together |
|---|---|---|
| 1. | 11, 20 and 49 | 16 times |
| 2. | 5, 11 and 13 | 16 times |
| 3. | 4, 10 and 25 | 16 times |
| 4. | 20, 33 and 43 | 16 times |
| 5. | 12, 30 and 32 | 15 times |
With 18,424 possible three-number combinations from 49 numbers, the expectation per triple is only about 5 times across all of lottery history. The top triple at 16 looks magical but is statistically unremarkable for the sample size: some triple has to be on top. Same conclusion: these numbers say nothing about the next draw.
Has the same 6-number combo ever been drawn twice?
We checked. Answer: never. 5.007 main draws since 1955, none of them with an identical six-number combo (without the bonus number).
Does that prove the drum has memory? No. There are 13,983,816 possible six-number combinations from 1 to 49. After 5.007 draws, only roughly 1 in 2.793 possible combos has ever shown up in a real draw. The vast majority of combinations have never been drawn at all.
Using a math trick called the birthday paradox you can calculate how likely it is that any two combos out of 5.007 draws would collide. Result: about 59%. Meaning: a collision would have been expected, but the fact that it hasn't happened yet is also well within the realm of pure chance. No evidence of memory. Just luck that we haven't seen it yet.
Does playing the same numbers every week increase my chances?
One of the stickiest myths at the local pub. Answer: no, not by a tenth of a percent. Whether you've played the same six numbers every week for 30 years or roll fresh ones every week, the chance of a six-match stays identical: 1 in 13,983,816.
The fallacy behind it: many believe "if I keep at it, my numbers will eventually come up". Mathematically false. The drum has neither memory nor a soft spot for regulars. Every draw is an independent event. The 5.007 draws so far don't change the probability of the next one.
What does change: your cumulative lifetime chance rises the more often you play, simply because you have more attempts. But that's true for any number combination, not specifically for "yours". Picking new numbers every week is statistically just as (un)successful.
One small caveat: if your favourite combo is unusual (no birthdays, no diagonal pattern, no consecutive sequence), you'd win with the same probability but share the jackpot in case of a hit with fewer co-winners. That's not a chance increase, it's a payout-share increase. More on that in the next section.
Can knowing this help you in the lottery?
Yes, but differently than most people assume. You can't increase your chance of winning. But you can optimize the expected payout if you do win. The trick is called payout optimization.
If a jackpot is up for grabs and you pick 1, 2, 3, 4, 5, 6 or 3, 6, 9, 12, 15, 18 or birthdays from 1 to 31, and win, you'll be sharing the pot with dozens of others who had the same "intuitive" idea. Sometimes with hundreds of others.
If instead you pick 14, 22, 31, 38, 41, 47 (deliberately unremarkable numbers, no pattern on the ticket, no birthdays), you win with the same probability, but if the jackpot lands, you split it with far fewer people. The expected value of your ticket doesn't go up because of the probability (still 1 in 139,838,160), it goes up because of the better quote in the win case.
Caveat: this doesn't change the lottery's negative-expected-value reality. Even optimally played, lottery is a long-run losing game. Here's the calculation with 70 years of data.
But what about lucky numbers, zodiac signs, and birthdays?
If the drum has no memory, a few popular theories fall with it. We already took apart "hot" and "cold" numbers above. Here are the next four big strategies, viewed through the same sober lens.
Birthdays and personal dates
By far the most common number-picking strategy: anniversaries, kids' birthdays, the date of the first meeting. It has its charm, because the ticket feels less like random chance and more like a life story. If you play that way, you have a reason that survives every math argument.
Statistically, it does not change your chance of winning. Birthdays all sit between 1 and 31, though, and that has two consequences. First: the numbers 32 to 49 are completely missing from the ticket, so a good third of the number range is out. Second: a huge number of other players pick the same way, so if you win, you split the jackpot with more people than average (see payout optimization above). The probability of a six-number hit stays at 1 in 13,983,816. The expected payout if you win drops noticeably.
Lucky numbers assigned to zodiac signs
Astrology apps and tabloid horoscopes like to hand out three or four "lucky numbers" per zodiac sign, sometimes derived from birth-date digit sums, sometimes from tarot or numerology, often just pulled out of a hat. The idea: certain numbers are supposedly especially favorable for your zodiac sign.
If you compare 70 years of draw data with an arbitrary zodiac-to-number assignment (for example, Aries equals 1 to 4, Taurus equals 5 to 8, and so on), the spread between the most successful and the weakest sign sits below two percent. Exactly within the corridor that standard deviation predicts for evenly distributed randomness. Put differently: any zodiac-to-numbers assignment would be statistically equally (un)successful, whether sorted by element, zodiac wheel, or favorite color.
Lucky and unlucky numbers like 7 and 13
The number 7 has been considered lucky since antiquity (seven wonders of the world, seven days of the week, seven dwarfs), while 13 is seen as unlucky in many countries. Some players deliberately pick 7, others consistently avoid 13. Both are understandable, because these meanings run deep culturally.
The data shows none of that. Both 7 and 13 sit inside the normal scatter range of all 49 numbers, with deviations from the average in the single-digit percentage range. The drum knows no numerology. It knows only 49 balls, all with the same weight, all rolling the same way.
Repeated digits and ticket patterns
11, 22, 33, 44. Diagonals across the ticket grid. All numbers around the edge. One full column. Visually pleasing picks that are easy to remember and fun to mark. No objection to the practice here either.
The drum does not see the ticket. It draws balls, not patterns. What matters is the combination, not the geometric arrangement. Repeated digits and ticket patterns have the same payout-optimization effect as birthdays, only stronger: thousands of other players pick exactly these noticeable patterns, so if the jackpot lands, it is split among many people.
To be clear: none of these strategies is dumb. If you pick birthdays for fun or put 7 on the ticket because it feels lucky, you are doing everything right, as long as you know it is fun and not optimization. But if you believe one of these methods increases your chance of winning, you are wrong. The jackpot hit odds do not care about zodiac signs, birthdays, or lucky symbols. They depend only on the math, and the math is mercilessly indifferent.
Take-aways in three bullets
- The drum has no memory. Past draws have zero influence on the next one. Every ball has exactly the same chance every time.
- Hot and cold numbers are illusions. What looks like a pattern is statistical noise that our pattern-spotting brain interprets as signal.
- If you do play: pick unremarkable numbers. Your chance of winning doesn't change. But in the (very unlikely) event of a win, you'd split the jackpot with fewer co-pickers.