How a domain trend forecast is made
Most watched topics never get a forecast. This is what has to be true before one is published, and what the published numbers mean.
The rest of this site is written to be understood without knowing anything about domains. These pages are for the reader who wants to know whether the method holds up before trusting anything it produces. Every figure quoted here is read from the code that uses it, so these pages cannot drift away from the engine they describe.
Most watched topics never get a forecast. This is what has to be true before one is published, and what the published numbers mean.
A track record is only worth reading if the failures are in it. This is how a call is settled and what stops the record being quietly curated.
A forecast is only as good as what it reads. This is what gets read, how much each kind counts, and how a bad source is taken out of the number.
Every figure on a topic page is either a measurement, a forecast, or a statement that something could not be measured. Telling them apart is the whole skill.
Prefer to see the method applied rather than described? The public record carries one page per graded call, and the trend board shows what is being watched right now.