Explainer
Hit frequency vs. RTP: two different measurements
The short answer
Hit frequency counts qualifying outcomes relative to the number of events. Return measures amounts. Both require clear definitions, and neither can be substituted for the other.
In this article
The word “hit” sounds self-explanatory until two reports use it differently. One might count any positive amount, another only an amount above the original input. Before comparing a percentage, identify what qualified for the numerator and what was counted in the denominator.
Count and size are separate columns
Illustrative example
| Sample | Nonzero outcomes | Sum of amounts returned | Nonzero frequency |
|---|---|---|---|
| A: four returns of 0.5 units | 4 of 10 | 2 units | 40% |
| B: one return of 8 units | 1 of 10 | 8 units | 10% |
Sample A has the higher frequency of nonzero outcomes and the smaller sum of returned amounts. Its observed return ratio is 2 divided by 10, or 20%. For B, that ratio is 80%. There is no contradiction: the percentages summarize different features of the samples.
Define the event before counting it
If “hit” means a return greater than the one-unit input, A records no qualifying events. If it means any positive return, A records four. The data stayed the same; the definition changed. A report should state how it handles partial returns, a return equal to input, and multi-step features.
The distinction between a count and its relative frequency is introduced in OpenStax. Applying it responsibly means retaining the raw counts. “40%” from four events out of ten gives the reader less evidence than the same proportion from a much larger, properly defined sample.
Observed frequency is not automatically a specification
A short observed sample describes what happened within that sample. To describe a configured probability, a publisher needs the relevant model or documentation. Rounding, omitted events, and a changed event definition can all make two displayed percentages appear more comparable than they are.
