What randomness is, and what it is not
A random process is one where you cannot predict a single result in advance, even though you may know the chances of each possible result. A fair coin is the classic case. You cannot say whether the next flip will be heads, but you know the chance is 1/2.
People often expect random sequences to look tidy, with results alternating and balancing out quickly. Real random sequences are lumpier than that. Understanding this makes it easier to avoid seeing patterns and causes where there are none.
Independence: chance has no memory
When events are independent, the outcome of one does not affect the next. Each flip of a fair coin, each roll of a fair die and each spin of a fair wheel starts afresh.
That is why the chance of heads is 1/2 on the flip after five tails in a row. The coin cannot know what happened before. The belief that a result is “due” after a run is called the gamblers fallacy. A related, opposite mistake is the idea that a run will keep going, and the comparison of the two is the topic of hot hand belief.
There is one exception to be careful about. If you are drawing from a shrinking pool, such as cards from a deck without putting them back, the events are not independent. The change in what is left alters the next chance, as covered under without replacement. A coin or a die does not shrink in this way.
Short runs look uneven
Take ten flips of a fair coin. There are 2 × 2 × … × 2 (ten times) = 1,024 equally likely sequences. How many of them contain exactly five heads? The number is 252, so the chance is 252/1,024, about 24.6%.
That means about three times out of four, ten flips do not split exactly five and five. Seven heads and three tails, or the reverse, is quite ordinary. A perfect balance is not the typical result, and it is not required for the coin to be fair. Counting the sequences like this uses the same ideas as the binomial distribution.
Getting all ten heads has probability 1/1,024, less than 0.1%. It is rare for any particular ten flips, but over a lot of people flipping, someone will see it. Rare events are expected somewhere if there are enough opportunities.
Streaks are normal
Because each flip is independent, runs happen simply by chance. In a long enough sequence of flips, a run of five or six of the same side is not a sign that anything is wrong. Our minds are tuned to notice patterns, so we tend to remember streaks and forget the many ordinary stretches.
This also explains why people see hot and cold spells in everything from games to sports. Some of those streaks can be real, as skill and form change, but chance alone produces a surprising number of them. When a sequence is generated by software, designers sometimes alter the pure randomness to avoid streaks that feel unfair, which is discussed in streak breakers. Computers also usually rely on a formula rather than true randomness, an idea explained under pseudo-random generators.
Short runs versus long-run averages
The long-run behaviour of random events is more regular than the short-run behaviour. Over many flips, the share of heads tends to settle near one half. That tendency is the law of large numbers.
There is a subtle point that often confuses people. As the number of flips increases, the proportion of heads usually gets closer to one half, but the raw difference between heads and tails does not necessarily shrink. Imagine that after 10 flips you have 6 heads and 4 tails, a gap of 2, and that after 1,000 flips you have 510 heads and 490 tails, a gap of 20. The gap grew, but the proportion went from 60% to 51%, which is closer to one half.
The law of large numbers is also not a correction mechanism. The coin does not make up for earlier imbalances; later flips simply dilute them as they are joined by more results. If you had ten extra heads at the start, the next hundred flips will not remove them. They just matter less in the average.
Why small samples mislead
Because short runs are so variable, judging a coin, a die, a player or a strategy from a few results is unreliable. A handful of outcomes can look impressive or terrible by luck alone. Trust grows with the number of observations, which is the lesson of sample size. A related effect is that unusually extreme results tend to be followed by more ordinary ones, as described in regression to mean. This is not because chance corrects itself, but because extreme results are partly made up of luck that does not repeat.
Practical lessons
- Do not expect randomness to look evenly spread. Clumps are normal.
- Do not treat a streak of independent events as a reason to change what you expect next.
- Be careful drawing conclusions from small numbers of results.
- Remember that an average over many trials says very little about what happens in a single session.
This last point matters for gambling. Games with a house edge lose money on average, and randomness can produce both winning and losing streaks along the way. A winning streak is no evidence of a method that works, and a losing streak does not mean a win is coming. Because outcomes are uncertain and gambling can be harmful, anyone who feels pulled to keep playing should consider contacting a local support service.