Proof standards
Definitions and publication rules for realtime.tips strike rate, LSP, BSP, ROI, confidence bands and settled selection records.
A proof total should be reproducible from the selection-level rows in scope. Winning and losing settled selections remain in the record, and pending, void or non-runner rows are labelled separately.
All-selections results are kept distinct from confidence-band or other filtered subsets. The method used to choose a subset is described next to it.
Strike rate is winners divided by settled selections. Level-stakes profit uses the same nominal win stake for every qualifying selection. SP and BSP returns are separate series, and BSP availability or commission treatment is stated.
ROI is profit divided by total stakes. Date range, settled count and price basis accompany the headline so a percentage is not presented without its denominator.
A band or rule selected after reviewing historical performance is not described as proven on that same history. It becomes a published hypothesis and is monitored on later selections without silently changing the definition.
Where the product changes model version, prompt, inputs or selection logic, the record identifies the change so readers can avoid treating unlike periods as one stable system.