Editorial standards
How realtime.tips researches, checks, updates and links AI racing articles to the underlying method and proof record.
Performance statements should name the model or system, date range, settled selection count, metric and price basis. A historical filter is labelled as historical until it has accumulated forward results.
We distinguish model output, calculated site data and external benchmarks. When a benchmark changes over time, the article records the date on which it was checked.
Each article must answer a distinct reader question, add an explanation that is not already covered by another page and link to the relevant product or proof surface without hiding limitations.
We do not create a new article for every date when one maintained evergreen page answers the query better. Event guides and monthly reviews are published only when they contain specific evidence.
Articles carry publication and modification dates. We update material numbers when the source record changes and retain the distinction between the original test window and later results.
A correction should improve the record, not erase an inconvenient losing period or rewrite a pre-race selection after settlement.