Lesson 3 · Independence

Past Draws Do Not Predict the Next

Results archives, frequency charts and “overdue number” lists all rest on one hope: that earlier draws say something about the next one. In a properly run lottery they do not. This lesson shows why, and then calculates how uneven truly random results should look, because that unevenness is what makes the charts so persuasive.

Any number, any draw

12.2%

6 ÷ 49 in a 6-from-49 game, every time.

The principle

A drum has no memory

Two events are independent when the result of one does not change the probability of the other. Lottery draws are built to work that way. The balls are mixed fresh for each draw, and the machine has no store of past results to consult.

Regulated lotteries take this seriously. Draw equipment and ball sets are tested, draws are usually overseen by independent observers or auditors, and many operators publish their procedures. The details differ by country and operator, but the aim is the same: each draw should be a fresh, unbiased selection.

Worked example

One number across two draws

P(7 drawn this week) = 6 ÷ 49 ≈ 0.122P(7 drawn next week) = 6 ÷ 49 ≈ 0.122P(7 drawn both weeks) = 0.122 × 0.122 ≈ 0.015

What it means: a number appearing twice in a row is unusual before the first draw happens, about 1.5%. But once it has appeared this week, the chance for next week is back to 12.2%. The rare part has already happened.

A famous mistake

The gambler’s fallacy

The belief that a result becomes “due” after a long absence is known as the gambler’s fallacy, or sometimes the Monte Carlo fallacy. The second name comes from an evening in 1913 at the Monte Carlo Casino, where a roulette wheel is reported to have landed on black 26 times in a row. As the streak grew, players are said to have bet ever more heavily on red, convinced it had to come.

On a European wheel, 18 of the 37 pockets are red, so each spin gave red a chance of roughly 48.6%. That was true on the first spin and on the twenty-sixth. The streak was extraordinarily unlikely when viewed from the start of the evening; it said nothing at all about the next spin.

Psychologists Amos Tversky and Daniel Kahneman described the underlying habit in their 1974 paper “Judgment under Uncertainty: Heuristics and Biases” in Science. People expect short sequences to look like the long-run average, so a run of black “should” be balanced by red soon. Chance does not owe anyone a correction.

Lottery charts invite the same error. A list showing that 31 has not appeared for 15 draws reads like a forecast. It is only a record.

Calculation

How many numbers look “cold” by chance?

The chance that a particular number misses one 6-from-49 draw is 43 ÷ 49. Missing several in a row multiplies:

Miss 10 draws: (43 ÷ 49)10 ≈ 0.27149 × 0.271 ≈ 13 numbersMiss 20 draws: (43 ÷ 49)20 ≈ 0.07349 × 0.073 ≈ 3.6 numbers

What it means: at any moment, around 13 numbers will typically have missed the last ten draws and three or four will have missed twenty, with nothing unusual going on.

Calculation

How uneven should 100 draws look?

Over 100 draws, each number is expected to appear 100 × 6 ÷ 49 ≈ 12.2 times. The usual spread around that average is the standard deviation:

√(100 × 0.122 × 0.878) ≈ √10.7 ≈ 3.3Typical range ≈ 12.2 ± 2 × 3.3 → about 6 to 19

What it means: one number turning up 18 times and another only 7 times over 100 draws is ordinary. A “hottest numbers” chart will always have a top and bottom, even for a perfectly fair machine.

The law of large numbers

Averages settle; gaps need not close

Jacob Bernoulli’s Ars Conjectandi, published in 1713, set out what is now called the law of large numbers: over many trials, the proportion of an outcome settles near its probability. It does not promise that earlier imbalances get paid back.

10 coin flips: 7 heads → 70% heads, 4 more than an even splitNext 1,000 flips (expected): 500 headsTotal: 507 of 1,010 → 50.2% heads, still 4 more than even

The share drifts toward 50% because the early surplus is diluted, not because the coin compensates.

A surprising count

Consecutive numbers are common

Many players avoid tickets with neighbours such as 22 and 23, because they look too tidy. The count disagrees. The number of 6-from-49 sets with no two consecutive numbers equals C(44, 6), a standard counting result.

C(44, 6) = 7,059,0527,059,052 ÷ 13,983,816 ≈ 0.505At least one consecutive pair ≈ 1 − 0.505 ≈ 49.5%

Roughly half of all draws contain a consecutive pair. And 1-2-3-4-5-6 is exactly as likely as any other set.

Render of a single plain brass coin lying flat on a graphite surface under soft light
A coin is the simplest independent event: each toss starts again at one half.
Coincidences

When the same numbers come up twice

In September 2009 the Bulgarian state lottery drew the same six numbers in two consecutive draws. The result was widely reported and prompted an official inquiry, which found no wrongdoing. For any one pair of draws, a repeat is as unlikely as a single line hitting the jackpot. But with many lotteries running many draws over many years, such coincidences are expected to happen somewhere eventually.

What it means: rare per draw is not the same as rare across the world.

Where choice still matters

What independence does not cover

Independence says your chance of matching the numbers is fixed. It says nothing about how many other people chose the same numbers. Studies of player choices in several countries have found that many people pick numbers from 1 to 31, the range of calendar dates, and favour visual patterns on the ticket. If such a set ever won, the prize would be split more ways. That affects the amount, not the probability, and it is covered in the average return of a ticket. For the counting behind these figures, see counting every possible ticket.