The statement
The law of large numbers says that as you repeat the same random trial more and more times, the average of the results tends to move closer to the true expected value. Flip a fair coin and the proportion of heads drifts toward one half. Play blackjack under fixed rules and the average result per hand drifts toward the game’s expected value, a small negative number for the player.
The law is about averages, not totals. This difference is the source of most misunderstandings.
How slowly it works
The tendency is real but slow. The typical gap between the observed average and the true value shrinks with the square root of the number of trials.
For a fair coin, the standard deviation of the proportion of heads is 0.5 divided by the square root of n:
- 100 flips: 0.5 / 10 = 0.05, so the proportion is typically within a few percentage points of 50 percent
- 10,000 flips: 0.5 / 100 = 0.005, typically within about half a percentage point
A hundredfold increase in flips tightens the gap tenfold.
For blackjack, the standard deviation per hand is a little over one unit; using 1.1 for illustration, the typical gap in the average result per hand is 1.1 / the square root of n:
- 100 hands: 0.11 units
- 10,000 hands: 0.011 units
- 1,000,000 hands: 0.0011 units
If the true edge against the player is, say, 0.005 units per unit bet, then after 100 hands the noise (0.11) is more than twenty times the edge. After 10,000 hands it is still more than double. Only around a million hands does the noise (0.0011) fall well below it. The average does converge, but it takes a volume of play few people ever reach.
What the law does not say
It does not say results even out. After a long losing stretch, the next hand is dealt with the same odds as before. Believing that a win is due is the gamblers fallacy. The law of large numbers works by dilution, not by compensation: early deviations are not reversed, they are simply swamped as more hands accumulate.
A small illustration: suppose you are 20 heads ahead of the expected count after 100 flips. If the next 10,000 flips are perfectly average, the total is still 20 ahead, but in proportion terms the gap is now 20 out of 10,100, about 0.2 percent, much smaller than the 20 percent it was after 100 flips (20 out of 100). The gap in count never closed, but it stopped mattering as a proportion.
It does not say totals converge. The average result per hand approaches the expectation, but the total profit or loss does not approach any fixed number. In a negative game the total loss tends to grow, in proportion to the hands played, and with it grows the absolute spread.
It does not guarantee any particular session. It says nothing certain about tonight.
At the table
The practical consequences are the following.
- A short run of good results is weak evidence about the edge. A player can win for a whole night under a house edge.
- A short run of bad results is not evidence of a rigged game or a cold table. Losing streaks of some length are an ordinary feature of random sequences.
- Judging a playing method from a handful of sessions is unreliable. A method that shows no real advantage can look excellent or terrible over a few hundred hands.
- A table of odds, like those built by simulation, describes long-run averages. It does not tell you what the next hour will bring.
Why casinos still rely on it
For the casino, the number of hands across all players is huge, so the average result comes close to the expected one, which is why the house edge is dependable for them. For one player, hands are few and variance dominates. The same law produces both experiences.
A balanced takeaway
Understanding the law is a defence against two common errors: reading too much into short runs, and expecting luck to turn in your favour. It also explains why the cost of playing is better described by the expected loss than by any single result.
Blackjack has a house edge, and nothing in this law changes that. Play only with money you can afford to lose, and contact a local support service if gambling is becoming a problem.