Why ratings wobble
A rating is meant to settle near a player’s true level. In practice it never sits still. Each game moves it by some amount, and results contain luck, so the number wanders up and down around where it would settle. That wandering is rating volatility in the everyday sense: the size of the short-run swings in a displayed rating.
This is different from the modelled volatility quantity in some formal systems, which describes how erratic a player’s results are in general. Here the focus is simply on how much a rating bounces and how to read it.
Streaks are normal in random sequences
Many players are surprised by how often streaks occur. Take a player who wins exactly half their games. The chance of winning five in a row is 0.5 multiplied by itself five times, which is 1/32, about 3 percent, from any given starting point. That sounds rare, but across hundreds of games there are many starting points, so a run of five or more wins or losses will usually appear somewhere. Longer runs of seven or eight also show up occasionally in long sequences.
This is the same pattern described under random clumping: random sequences contain more clusters than intuition expects, because people tend to expect results to alternate. Streaks by themselves are poor evidence that something changed.
How a streak moves a rating
The size of a rating change from a streak depends on the system.
- In a basic Elo setup with a K-factor of 32, a run of five wins against evenly matched opponents is worth roughly 16 points each at the start, so about 70 to 80 points in total, slightly less as the rating rises and the expected score of each game goes up. A five-game losing run subtracts a similar amount.
- A higher K gives bigger swings and a lower K smaller ones.
- Systems tracking rating uncertainty move faster when confidence is low, such as early in a season.
- Some ladders add streak bonuses or hidden multipliers, which amplify the movement beyond what the results alone justify.
The consequence is that a player with a completely unchanged skill level can see their rating shift by dozens of points over a single evening.
Separating real change from noise
The core difficulty is that improvement and luck can look identical in the short run. A few guidelines help.
Look at longer windows. A rating graph over weeks says much more than one over hours. A real improvement usually shows as a sustained shift in the average level, while noise tends to return to where it started.
Compare against the expected range. If your results are very near 50 percent over a few hundred games, a single hot week is not unusual. A rough rule: the more games, the narrower the range of plausible luck, and the more weight a difference carries. This is the practical content of the sample size idea.
Check the quality of opposition. A climb through a stretch of weaker opponents is less informative than one against peers or stronger players. Matchmaking variation, including the queue time tradeoffs that sometimes produce uneven games, affects how much a run of wins really shows.
Look at performance, not only outcomes. In team games a win can come from teammates, and a loss from teammates too. Reviewing your own play through replay review tells you more about whether your level changed than the result alone, which fits the approach of results thinking: judge decisions by how good they were, not simply how they turned out.
Look for a cause. A genuine improvement usually has an explanation, such as a change in practice routine, a new role or a better grasp of a matchup. A rating jump with no identifiable reason is more likely to be noise.
Wait for the pattern to repeat. If a new level persists across several separate sessions and several tens of games, that is more convincing than any single run.
The psychology of streaks
Streaks also change behaviour in ways that can be unhelpful.
- After a winning run, players may feel unbeatable, and the overconfidence bias can lead to riskier or sloppier play.
- The hot hand idea, that success makes further success more likely, is intuitively appealing, though in many settings it is weaker than people feel.
- After a losing run, emotional tilt can set in, making further losses more likely through poorer decisions rather than any change in skill.
- Loss aversion makes a drop in rating feel worse than an equal rise feels good, which tempts players to keep queuing to win it back.
Noticing these effects matters because they can turn a normal swing into a larger one. A decision to stop after a run of losses is often a way of protecting your average from your own reactions.
What systems do about volatility
Designers use several levers. Lower K or reduced uncertainty makes the displayed number calmer. Hiding the exact rating behind tiers makes small swings invisible, though it exaggerates movement near boundaries. Some games add rank protection so that one loss does not drop you a tier. Others deliberately speed up movement for players on strong streaks, assuming that the streak signals a mismatch between rating and skill. That assumption is sometimes right, particularly with smurf accounts or fast-improving beginners, but it can also reward luck.
A sensible way to read the number
Treat a rating as a noisy instrument. Expect it to move by a visible amount in either direction even if nothing has changed. Draw conclusions from patterns over many games, not from last night’s session, and remember that a single streak, up or down, says very little about how good you are.