The Law of Large Numbers: Why the Casino Wins Over Time

The law of large numbers is the reason a 2.70% house edge turns into reliable casino profit. Over a handful of bets anything can happen. Over thousands, your average result per bet is pulled steadily toward the expected value, and for every casino game that expected value is a loss. The casino does not need to be lucky. It only needs volume.

What the law actually says

In plain terms: as the number of independent trials grows, the average outcome per trial gets closer and closer to its expected value.

The law is about averages, not totals, and that distinction is where most misunderstandings start. Flip a fair coin and count heads. The typical gap between your count and exactly half is one standard deviation, √(N × 0.25):

• 100 flips: typical gap of 5 heads, which is 5% of the flips.
• 10,000 flips: typical gap of 50 heads, which is 0.5%.
• 1,000,000 flips: typical gap of 500 heads, which is 0.05%.

The absolute gap grows, from 5 to 500. The percentage gap shrinks, from 5% to 0.05%. Both facts are true at once, and both matter at the casino.

Averages converge, but totals never even out

Suppose your first 10 flips are all heads. You are 5 heads above the expected count of 5.

Now flip a million more times. The coin has no memory, so those million flips are expected to split evenly. You finish, on average, still 5 heads above half. Nothing paid back the early streak.

What changed is the proportion: 5 extra heads out of 1,000,010 flips is a rounding error. The law of large numbers works by dilution, not by compensation. Believing in compensation is the gambler's fallacy.

Your side of the table: $10 on red, over and over

A $10 bet on red in European roulette wins with probability 18/37. Its expected value is −$10/37 ≈ −$0.27, and its standard deviation is about $10, because every bet ends roughly $10 up or $10 down.

The expected loss grows in step with N. The swing grows only with the square root of N. Early on the swing dominates; later the loss does.

Over a short session, losing and winning nights both happen often. That is real, and it is what keeps the game fun. But by 10,000 bets, about 286 hours at 35 spins an hour, being ahead has become a one-in-300 event.

Volatile bets slow the law down, but never stop it

How fast the average settles depends on how spread out each bet is. Swap the $10 red bet for $10 on a single number. The expected value is identical, −$0.27, but the standard deviation jumps to about $58.38, because results are now either +$350 or −$10.

The single-number player stays in the game far longer, because one or two extra hits swing the total by hundreds of dollars. The two-standard-deviation crossover moves from about 5,472 bets to about 186,600 bets, more than 5,000 hours of play.

The expected loss is the same $0.27 per bet either way. Volatility changes how long luck can hide the edge, not whether the edge is there. The same logic explains why high-volatility slots feel so unpredictable; see RTP vs. volatility.

The casino's side: millions of bets

Now look at the same table from the other side. The casino is not one player making a hundred bets. It is the counterparty to every bet from every player, all day, every day.

Take 1,000,000 bets of $10 on even-money European roulette. The casino's expected win is $270,270, and its standard deviation is only about $9,996. Roughly 95% of the time, its result lands within about $20,000 of $270,270. For the casino to lose overall, luck would have to deliver a result 27 standard deviations below average, which for practical purposes never happens.

That is why a casino can treat its income almost like a utility bill. Individual tables lose on some nights, and a few very large bets can swing a short reporting period. But across the whole floor and the whole year, revenue tracks the house edge times the total wagered. The guide to how casinos make money shows how they forecast it.

The law of averages is not a law

People often say a losing streak means wins are coming because of the law of averages. There is no such law. The mathematical one says nothing about the next spin, and it never promises that your total will return to even.

Short runs are wild because they are short. In 100 coin flips there is about an 81% chance of a streak of six or more identical results somewhere in the sequence. Watch it happen in the free coin flip lab, then notice that the running percentage of heads still drifts toward 50% as the flips pile up.

What this means for you

• Short-term luck is real. Over an evening, winning is common. That does not mean the game has a weakness.
• Long-term results are not luck. The more you play, the more your total looks like the house edge multiplied by everything you wagered.
• Your cost is controllable. Lower-edge bets, smaller stakes and fewer bets all shrink expected loss. The house edge calculator shows the numbers for your game and pace.
• Swings need a buffer. Because short sessions vary so much, size your budget for the swing, not just the average. The bankroll and risk of ruin guide covers how.

The bottom line

The law of large numbers turns uncertainty into near certainty, but only for the average. Over a few bets, your result is dominated by luck. Over thousands, it is dominated by the house edge, because expected loss grows with N and luck grows only with √N.

The casino lives in the long run; you visit it. If you play, play short, play small, and treat the expected loss as the price of the evening. If the losses start to feel like something to win back, read the responsible gambling guide.