Recorded meetings and scored meetings are different
A matchup may have 30 recorded results but only 18 meetings with usable point totals. A win/loss summary and a point-total summary can legitimately have different denominators. Use the sample attached to the metric you are reading.
As an illustrative example, suppose the available totals are 70, 80, and one missing score. The average of the two scored meetings is 75. Treating the missing score as zero would incorrectly produce 50. No available scores should mean unavailable, not a measured zero average.
A recent window is not a complete archive
A selected 20-meeting window means up to 20 usable observations from the available publication. It does not promise 20 records, an exhaustive archive, or coverage of the most recent 20 fixtures: newer fixtures may have a result but no usable point scores.
A current board is a selection of active matches. A player missing from that selection is not proof they are inactive, and a matchup missing from the board is not proof no history exists.
Small samples can move quickly
In five matches, changing one observation changes a hit rate by 20 percentage points. In twenty matches, it changes the rate by five percentage points. That is why the count belongs next to the percentage.
A larger sample does not resolve every uncertainty. Older results may describe different circumstances, and repeatedly playing the same opponents can make observations less independent than they appear. Compare windows without treating either as a guaranteed forecast.
Which timestamp should you check?
A board's updated time describes publication. A scheduled time describes when a fixture is expected to start. A latest-meeting date describes a historical observation. Those timestamps answer different questions.
A newly refreshed board can still contain old history. Check the most recent dated result and scored observations before describing a statistic as current. Feed delays, rescheduling, and corrections can change what is displayed.
Citing a statistic accurately
Include both players, league, selected line, scored count, window, and available dates. Link to the exact match page from the current board so another person can inspect the context. If the match is absent from the current publication later, that page may be unavailable; it does not establish that the historical match never existed.
For example: “In the displayed sample of 10 scored meetings, 6 totals were above 74.5.” That describes the evidence more accurately than assigning the next match a 60% chance of going over. This example is illustrative, not a claim about a current fixture.