Chance is a good explanation until it is not
Luck is the easiest explanation for a surprising result, and often the right one. Streaks, upsets and strange runs of drops all happen in purely random systems. But chance is not an all-purpose answer. It predicts a certain range of results, and when an outcome falls well outside that range, it is worth asking what else is going on.
The skill is in knowing where the edge of that range lies.
Measuring how far chance can reach
For a win rate, the typical spread caused by luck alone shrinks with the number of games. If two evenly matched players meet, each has a 50 percent chance per game, and the standard deviation of the observed win rate is 0.5 divided by the square root of the number of games.
- After 100 games, that is 5 percentage points.
- After 400 games, it is 2.5 percentage points.
A common rough rule treats results more than about two standard deviations from expectation as unlikely to be pure chance. For 400 games between equal players, two standard deviations is 5 points, so a 56 percent win rate sits beyond that line, while 52 percent does not. With only 100 games, two standard deviations is 10 points, so even a 58 percent record could come from nothing but luck.
This is a rule of thumb, not a verdict. About one in twenty equal players would still land outside two standard deviations on luck alone, and the more comparisons you make, the more of those you will find. But the main message holds: a small edge only becomes visible after many games, and a large edge becomes visible quickly.
What takes over when chance runs out
When luck does not explain a gap, several other causes are usually worth checking, in roughly this order.
Skill and preparation. A real, repeatable advantage in decision-making, execution or knowledge produces results that persist. The test is whether the edge shows up again in new data, not just in the games that first drew attention. Questions about luck and skill are explored in more depth elsewhere, but the key sign is persistence over time.
Matchup and context. A result may reflect who was played, not how well. A player who faces weaker opponents will post better numbers. A team strong on one map may look weak overall if the pool is mixed. Head-to-head records and map pools often explain results that look like luck or like skill but are actually context.
Hidden game mechanics. Not every random-looking system is plain random. Some games adjust odds after a failure, and the details of pseudo random distribution mean that streaks of bad luck are shorter than pure chance would give. In such systems, assuming independent rolls can lead to wrong conclusions about “unlucky” or “lucky” outcomes.
Dependence between events. The simple formulas assume that trials are independent. In real games they often are not. Winning an early fight changes the likelihood of winning the next one. Confidence, momentum and resources all link events together. When results are linked, runs are longer and spreads are wider than the independent model predicts.
Bias in what you see. People remember striking events and forget ordinary ones. They also pick out a result because it looked unusual, which inflates how rare it seems. If you look at enough streamers, players or items, some will have extraordinary records by luck alone.
Measurement problems. Cheating, bugs, unreported rule changes and misread data can all produce odd results. These are less common than ordinary variation, and they should be a later explanation, not the first one.
The multiple-comparisons trap
When many things are checked, something unusual will appear. If you examine twenty independent events, each with a 5 percent chance of looking suspicious, the chance that at least one does is 1 minus 0.95 to the 20th power, which is about 64 percent. So finding one “impossible” streak among twenty players does not by itself show that anything is wrong.
A safer approach is to decide in advance what you will test and to treat after-the-fact discoveries as ideas to check on fresh data. This is the same principle behind caution about any claimed pattern exploitation based on a handful of results.
Where the edge of chance sits in design
Designers know that pure randomness has limits as a feature, and they shape it. A few examples of rules that deliberately limit the reach of chance:
- Caps and floors on damage ranges so that no single roll is overwhelming.
- Mechanisms that guarantee a result after a number of failures, which bound how unlucky a player can be.
- Series formats that lengthen the sample before a winner is declared.
- Matchmaking that pairs similar ratings so that skill differences do not swamp everything else.
Each of these accepts some randomness for drama and then bounds the damage it can do. They are also reasons why raw results from a game should be interpreted with the rules in mind.
A short method for borderline cases
When a result seems too good or too bad to be luck:
- Count the trials. A small number usually means luck is a fine explanation.
- Estimate the spread luck would produce with that count, using the square-root rule.
- Check how many similar results you looked through before picking this one.
- Ask whether the system uses hidden adjustments that change the odds.
- See whether the pattern holds up in data you did not use to find it.
If it passes all five, a non-random explanation becomes much more plausible. If not, luck remains the simpler answer.
Chance and certainty are both limited
It is tempting to treat luck as a last resort for explaining what we do not understand, or as an excuse. Neither is quite right. Chance has real limits: it cannot easily produce large, persistent, repeatable gaps. But it also reaches further than intuition suggests, producing long streaks and lopsided short records regularly. Understanding both sides lets you hold conclusions loosely when the evidence is thin and more firmly when the sample is large.