Casino games are games of chance with a built-in house edge. No strategy, system or timing changes the odds in the player’s favor over time, and most players lose money over time. Nobody should gamble money they cannot afford to lose. This article explains a statistical principle; it does not describe a way to beat the odds, and nothing in it predicts any one session.
What the law says
The law of large numbers states that as the number of independent trials grows, the average result tends to move toward the expected value. Flip a fair coin a few times and the share of heads can be far from one half. Flip it thousands of times and the share is typically very close to one half.
Two parts of that sentence are easy to miss. The law concerns the average or proportion, not the raw count. And it describes a tendency across many trials; it says nothing definite about the next flip or any short run.
Proportion versus count
Take a fair coin. The expected number of heads in n flips is n/2, and the standard deviation of the count is √n ÷ 2.
| Flips | Expected heads | Standard deviation of the count | Standard deviation of the share of heads |
|---|---|---|---|
| 100 | 50 | 5 | 5 percentage points |
| 10,000 | 5,000 | 50 | 0.5 percentage points |
Check the pattern. Going from 100 to 10,000 flips multiplies the number of flips by 100. The spread in the count grows from 5 to 50, a factor of 10, which is the square root of 100. The spread in the share shrinks from 0.05 to 0.005 because the count’s spread is being divided by a much bigger total. So the gap in heads usually grows in absolute terms while the proportion gets tighter. A common mistake is expecting the raw difference to shrink. It does not.
It is not a force that corrects past results
A frequent misreading, sometimes called the “law of averages,” is that if a coin has landed heads too often, tails is now due. The law of large numbers does not say that. Coin flips are independent and have no memory. Early imbalances are not canceled out; they are diluted, because new trials pile up around the average and the early fluke becomes a smaller share of the total.
Suppose after 100 flips the count is 60 heads and 40 tails. The next 9,900 flips are expected to split evenly, 4,950 each. The final total is 5,010 heads, which is 50.1% heads. The 20-flip imbalance has not been repaid, it has faded in relative terms. This is the opposite of the reasoning in the gamblers fallacy, and the mistaken belief behind the idea that machines are “due.”
Applying it to a casino game
Consider a $1 bet on one number on a single-zero wheel with a 35 to 1 payout. The expected result is −1/37, about −2.7 cents per bet.
- A player making 100 such bets has an expected result of about −$2.70, but a standard deviation of roughly $58. One player might be well ahead, and another well behind.
- A venue taking a million such bets (hypothetical) has an expected result in its favor of about $27,000 on that volume, with a standard deviation of 5.84 × √1,000,000 = about $5,840. The spread is a small fraction of the expected amount, so the venue’s outcome is dependable by comparison.
The expected figure grows in proportion to the number of bets (1,000,000 × 0.027), while the spread grows only with the square root (√1,000,000 = 1,000 times the single-bet figure). The ratio of spread to expected result therefore shrinks as volume rises. This does not mean any particular figure will occur; it means the average result per bet becomes very close to the theoretical one.
That is the sense in which the law favors the house. The edge per bet is small and the same for each bet, and a venue combines the outcomes of an enormous number of independent bets from many people. The reported fraction of wagers retained, discussed under hold percentage, then lands close to the built-in figure. The venue still does not know which player will win on a given day.
What it means for an individual
The law of large numbers favors whoever has the positive edge and the large volume. A player faces the opposite combination:
- The edge is negative.
- The number of bets is small compared with the venue’s total, so the swings are large relative to the average cost.
Short-term wins are real and common because the spread is wide in relation to the edge. That is exactly why they mislead. Someone who is ahead after a few hundred bets is seeing noise that the edge has not yet overcome, not evidence that the game is beatable. Continued play pushes the expected result further in the wrong direction as the number of bets grows, and a bankroll can run out first.
Common misunderstandings
- “I’m due a win.” Independent trials have no memory.
- “Long-run averages will rescue me.” The long run is what accumulates the edge against the player.
- “The law means every session ends near average.” It does not. Single sessions remain unpredictable.
- “The average is guaranteed for me.” The average describes the pooled long-run results of enormous numbers of trials, not what any one person will experience.
Takeaway
The law of large numbers explains two facts at the same time. A game can feel random and generous for an hour, and a venue’s overall figures can still be remarkably steady. Understanding both makes it easier to recognize short-term luck for what it is, and to treat gambling as entertainment that costs money, with a budget of spare money, a time limit, and a willingness to stop. Anyone who finds that hard can reach out to a local support service.