The average describes the long run
Probabilities and expected values describe what happens across many repetitions. A single session is a small sample, and small samples are noisy. A fair coin gives 50% heads in the long run, yet a ten-flip evening can easily show seven or three. Nothing is wrong with the coin; this is what randomness looks like at small scale.
Understanding how much wandering is normal helps you avoid two common errors: deciding a game is rigged after a bad night, and deciding you have found a winning pattern after a good one.
How far is normal
The number of successes in a fixed number of independent tries has a typical spread of the square root of n × p × (1 − p), where p is the success chance.
Twenty attacks at a 55% hit chance. Expected hits: 20 × 0.55 = 11. The spread is the square root of 20 × 0.55 × 0.45, which is the square root of 4.95, or about 2.2. So a result of 9 to 13 hits is within about one standard deviation, and quite ordinary. Expecting exactly 11 is the unusual hope.
One hundred coin flips. Expected heads: 50. The spread is the square root of 100 × 0.5 × 0.5 = 5. A result from 45 to 55 is typical, and 40 to 60 (two standard deviations) is still fairly common, roughly 95% of the time.
The pattern in both cases is that the typical gap in count is a few units, which is a large share of a small total. As the total grows, the gap grows too, but more slowly, so the percentage error shrinks. This is the law of large numbers seen from the other side.
Rare events and droughts
Droughts are more common than intuition suggests. If a 5% event, such as a critical hit, is being rolled for, the chance of seeing none in 20 tries is 0.95^20, which is about 0.36. More than one session in three will go through twenty rolls without it. Similarly, the chance of at least one is only about 64%, so even a “likely” event is far from guaranteed.
For a 10% chance, the chance of nothing in 10 tries is 0.9^10, about 0.35. These dry spells are not evidence of a hidden penalty. Compare them with pity timer systems, which exist partly because players feel such droughts strongly even when they are statistically routine.
The chance of a long sequence of misses can be calculated with the multiplication rule for independent events, and the result is usually larger than people expect.
Streaks are normal
In 100 fair coin flips, a run of five or six identical results in a row is common, and the longest run is often around six or seven. A sequence with no streaks at all would actually look suspicious. People tend to expect random sequences to alternate more than they do, and so they read streaks as meaningful. Whether a streak signals anything is a question taken up in the hot hand belief, and for fair processes the answer is that it does not.
What short sessions can and cannot tell you
A short session can tell you about your own experience: you had a good or bad evening. It tells you little about the underlying odds, your skill, or whether a strategy works. A few points follow.
- Judge decisions by their reasoning, not the last result. A sound choice with a 70% chance of working still fails three times in ten.
- Be careful with small-sample statistics. A player who looks brilliant or terrible over a handful of matches is often just in the normal spread. Player statistics need a decent volume of data before they mean much, and extreme early results tend to move back toward the middle later, a pattern known as regression to the mean.
- Do not read patterns into noise. Sessions are not tied together by hidden balancing. A bad night does not make a good night more likely, and the gamblers fallacy is the belief that it does.
- Keep records honestly. Common mistakes when tracking your own results include recording only the memorable sessions and stopping at a convenient high point.
Money and risk
With real money at stake, short sessions matter even more, because swings are large compared with a modest bankroll. In games with a negative expected value, a lucky session can happen, and it does not mean the long-run expectation has changed; continuing to play tends to move results toward the average, which is a loss. Results are uncertain, so set a limit you can afford beforehand, and if play stops being enjoyable or starts causing stress, take a break and consider talking to someone.
A fair expectation for tonight
If you know a game’s average and its spread, you can set a realistic range for a short session: not “I should get exactly the average”, but “somewhere within a couple of standard deviations is normal”. When your result lands inside that range, there is nothing to explain. Only results well outside it, repeated across many sessions, give a good reason to question the assumed odds.