Communicating probability and harm: denominators, averages and limits

Put the denominator beside the percentage

A statement that an event occurs in 2% of observations is incomplete without knowing what an observation is. It might mean people, sessions, devices, days or recorded transactions. The time frame also matters: a daily proportion and an annual proportion cannot be compared as though they describe the same exposure. State the population, inclusion rule and observation period in the same sentence as the number. If these are unavailable, describe the limitation directly rather than giving the percentage a precision that the underlying record does not support.

Show absolute and relative change together

Consider a purely hypothetical comparison in which an event count changes from 100 among 10,000 observations to 200 among another 10,000. The proportions are 1% and 2%. The absolute difference is one percentage point; the relative increase is 100%, or a doubling. Both statements can be correct, but either alone can shape a very different impression. These invented counts demonstrate arithmetic and are not prevalence estimates. They also do not prove that any intervention caused the difference: comparability and study design would need separate examination.

Compare ways to describe the same example

A responsible caption includes both original proportions and the shared denominator. If the groups have different sizes, report their actual counts rather than implying that the same number of events occurred. Rounding should not turn a small nonzero probability into a claim of impossibility or a high probability into certainty.

Do not turn an average into a personal promise

The Gambling Commission explanation of RTP distinguishes a long-run average from a return on each individual occasion. The same general communication problem appears when an average waiting time is presented as a guarantee for every visitor. OpenStax’s expectation discussion provides the mathematical distinction between a distribution and its mean. Explain what the average summarises and what it leaves out, including variability and unequal consequences. A percentage describing aggregate output cannot establish an individual entitlement or erase previously incurred losses.

Separate mathematical properties from human consequences

The World Health Organization’s gambling fact sheet describes harms that can include financial distress, relationship damage and health effects, and notes that effects can extend beyond the participant. This is population-level context, not a diagnosis of a reader. A process being random or following a stated probability model does not logically establish that participation is harmless. Similarly, correcting a probability misconception does not eliminate social, financial or health consequences. Keep technical claims about a model separate from claims about wellbeing or acceptable exposure.

Avoid false certainty from labels and absence of evidence

Words such as fair, random, secure and safe refer to different properties and require different evidence. A random outcome mechanism does not prove secure data handling, and a published percentage does not prove regulatory approval. This website does not claim certification or licensing from the sources it cites. When a study fails to detect an effect, ask whether its design and sample could detect effects of practical importance. Absence of a statistically clear finding is not automatically proof that no meaningful consequence exists.

Explain uncertainty without shifting responsibility onto the reader

A clear explanation identifies what is known, what is assumed and what remains uncertain. It should not imply that a person experiencing harm simply failed to understand mathematics. Population conditions, product design and social circumstances can matter alongside individual beliefs. Avoid persuasive phrases such as guaranteed recovery, inevitable balance or risk-free participation. When discussing an illustrative model, keep the example abstract and avoid presenting it as a plan for choosing stakes or extending activity until a target result appears. Precision should reduce misleading reassurance rather than create a more sophisticated version of it.

Risk-communication checklist

• Name the event, population and time frame.
• Display counts and denominators.
• Pair relative changes with absolute differences.
• Label hypothetical examples clearly.
• Distinguish averages from personal outcomes.
• Keep randomness, security, legality and safety separate.
• State uncertainty without blame or recovery promises.

Before publishing, ask whether a reader could reasonably mistake the sentence for a guarantee. If so, move the limiting condition closer to the number. A well-written uncertainty statement remains specific: it identifies the missing evidence or model assumption rather than relying on a vague disclaimer at the bottom of the page.

Sources and verification

• Return to player: interpreting the average — Gambling Commission, Great Britain — 2026-10-23
• Mean, expected value and standard deviation — OpenStax, Rice University — 2026-10-23
• Gambling fact sheet, 2 December 2024 — World Health Organization — 2026-10-23