What the source says
How to interpret the signal
A ranking weight is applied to a predicted outcome for a particular viewer. It should not be read as a public exchange rate between raw likes, replies, or reports. The useful mental model is a portfolio of possible outcomes shaped by viewer context.
A practical example
A technical post may earn fewer raw reactions than a broad joke while still matching the intended audience more closely. Optimize the promise, evidence, and audience fit together instead of trying to maximize one visible counter.
Action for builders
Before publishing, ask whether the opening earns attention honestly, the body rewards the click, the topic fits the intended audience, and the post avoids a likely negative response such as mute or report.
Evidence boundary
This page explains the public repository and official documentation as verified on 2026-09-29. Production experiments and configuration can change, so no recommendation here guarantees distribution.