# Law of Large Numbers

> Why the average of many tries settles near what you expect. Why a coin never owes you heads, why wild scores drift back, and how chess rates skill.

**Course:** [Probability and Statistics](https://learn.benibix.com/learn/general-math)  
**Time:** 12 min

**After this lesson you can:**

- Explain the law of large numbers
- Spot the gambler's fallacy
- Expect a wild result to drift back to normal

## Why do averages settle down over many tries?

The Law of Large Numbers says that the more times you repeat a chance event, the closer the average gets to what you expect. Flip a coin 10 times and you might get 7 heads. Flip it 10,000 times and the share of heads will very likely sit between 49 and 51 percent. The noise does not vanish, but it shrinks next to the total.

This is why casinos and insurers can plan. Each bet or each driver is a gamble. Millions of them together are close to certain. It is also why a test of a few people tells you little, and a test of thousands tells you much more.

The Gambler's Fallacy is thinking a coin owes you heads after a long run of tails. It does not. Each flip has a 50 percent chance, whatever came before. The coin has no memory. In 1913, the ball at one casino landed on black 26 times in a row, and players lost fortunes betting on red.

The law of large numbers does not work by evening things out. A run of tails is not cancelled by later heads. It is simply swamped by thousands of new flips, which drown out its effect on the average.

Regression to the mean is the pull of extreme results back toward the average. A score far above your usual is part skill and part luck. The next time, the luck is unlikely to be as good, so the score tends to be lower. Nothing has to change for this to happen.

It fools people all the time. Flight teachers once noticed that praise after a great landing was followed by a worse one. Criticism after a poor one was followed by a better one. They decided criticism works. In fact, both results were just drifting back to normal.

The Elo Rating System changes wins and losses into a skill number. A player rated 200 points higher is expected to score about 0.76 per game. Beat a weaker player and you gain a few points. Lose to one and you drop many. Over many games, the rating settles near true skill.

![Four panels. Law of Large Numbers: a line for the share of heads in coin flips jumps around at first, then settles close to 50 percent as the flips pile up. Gambler's Fallacy: five heads in a row, and the next flip still shows a 50 percent chance of heads. Regression to the Mean: a first test score far above the average, and a second score closer to the average. Elo Rating System: a player rated 1600 meets one rated 1400, with an expected score of about 0.76 and 0.24.](https://learn.benibix.com/diagrams/gen6-chance-in-practice.svg)

### Example: A new coach after a bad season

A football team has its worst season in 20 years. The owner fires the coach, and the next season is much better.

Fans credit the new coach. But a team that bad was likely to drift back toward its normal level anyway, coach or no coach.

**Words in this lesson:** [Law of Large Numbers](https://learn.benibix.com/term/general/law-of-large-numbers), [Gambler's Fallacy](https://learn.benibix.com/term/general/gamblers-fallacy), [Regression to the Mean](https://learn.benibix.com/term/general/regression-to-the-mean), [Elo Rating System](https://learn.benibix.com/term/general/elo-rating-system)

**Question:** A fair coin has landed tails 8 times in a row. What is the chance of heads next?

**Answer:** Exactly half. The coin has no memory. Each flip is a fresh 50 percent chance, whatever came before.

---

Source: https://learn.benibix.com/learn/general-math/law-of-large-numbers
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-19
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
