The same average can hide very different outcomes

Average and spread are separate ideas

Averages compress information. They do not reveal how frequently awards occur or how large they are. A volatility description concerns that spread, although a provider’s “low” or “high” label need not use the same scale as another provider’s.

The UK Gambling Commission’s performance-monitoring guidance explains why volatility affects the range of observed returns. We use that relationship to distinguish concepts, not to rate any real game.

Two invented models with a 90% mean

These are teaching models, not real games. In both cases, multiply each award by its probability and add the results: the mean is 0.90 units for a one-unit cost. Model B has a much wider spread and a higher chance of a zero award on a single trial.

This constructed example shows why the same theoretical RTP cannot tell you the full outcome distribution. It makes no claim about any provider or how an actual product is configured.

A short sample cannot establish the design

A handful of rounds can look quiet or unusually eventful by chance. A description such as “high volatility” is not a schedule that says when a large award will arrive.

For independent random trials, a losing sequence does not increase the next trial’s win probability. Do not turn the idea of long-run averages into an assumption that a shortfall must be repaid during your session.

Compare only what is documented

This library provides no volatility ratings for real titles. Without an attributable method and the correct game version, a precise ranking would imply evidence we do not have.

• Check whether a volatility label has a stated method or scale.
• Keep theoretical RTP, hit frequency and maximum award in separate fields.
• Treat undated screenshots and short demonstrations as limited evidence.