Card rules · 2026-10-29
Why the Same Practice Results Can Show 40% or 50%
A percentage needs a denominator. Without one, a result-history summary can be numerically neat and still answer the wrong question.
A percentage needs a denominator. Without one, a result-history summary can be numerically neat and still answer the wrong question.
One log, two calculations
Suppose a teaching log contains ten entries. Two were cancelled, while four completed entries carry the label A. A accounts for 4 out of all 10 entries: 40 per cent. Among the eight completed entries, it accounts for 4 out of 8: 50 per cent.
Neither calculation repairs the other. They describe different populations. The first includes cancelled entries in its denominator; the second excludes them. The choice must match the question being asked.
Keep categories explicit
An incomplete entry is not automatically a completed non-A result. A replay of the same hand is not necessarily another hand. If a log combines these categories, counting rows can differ from counting distinct completed events.
Choose the denominator before calculating. If the question concerns completed events, cancelled entries stay outside that denominator but remain visible in the source log. Excluding them from one calculation is different from pretending they never existed.
These are descriptive calculations. They neither establish the mechanism producing outcomes nor predict the next result. A history percentage becomes more useful when its limits are visible, not when it is presented as a promise about what must happen next.
Ask whether a new row changes the denominator
Add a duplicate display of one completed entry to the ten-row example. Counting rows now gives eleven, but the number of distinct events remains ten. Deduplicate by the event identifier before calculating either percentage. Keep the duplicate in the evidence archive if useful; excluding it from the count does not require deleting the record of what was displayed.