How a finding earns the word “finding”.
Most sports analysis presents inference at the confidence of evidence. The method below exists to make that impossible to do quietly.
Three separated fields
Every claim we publish is written in three parts, kept visibly distinct:
Anchor
The raw observation and its source: the dataset, the sample size, the date range. No interpretation.
Observation
What the anchor shows on its own terms, stated without reaching for a cause.
Finding / Test
The interpretation, plus the specific condition that would show it to be wrong.
Inference cannot be promoted to evidence without visibly moving between fields. A reader who trusts nothing else can audit any claim by reading only the middle one.
Chronology before memory
Seasons are read forward, in order, rather than sampled backward from whatever outcome is already known. Reading backward from a result finds a cause every time, which is exactly why it proves nothing.
The contradiction register
Reports require a correction history. If new evidence changes a finding, the report must identify the original claim, new evidence, date and revised assessment. No live findings or public correction register are claimed until they exist.
What we do not do
- We do not sell betting selections or “plays”
- We do not accept, place, or broker wagers
- We do not publish win rates, records, or profit claims
- We do not describe any outcome as guaranteed, locked, or safe
Injury-impact reports
Report official availability, workload, replacement context, sample size, estimate uncertainty and a revalidation time. See the format.