Short enough to argue with, which is the point. If you think a number is wrong, everything you need to say so is on the film's own page.
We read the review, not the star rating.
A rating with no sentence under it gives us the temperature and nothing else. Where a critic published their own rating we carry it straight across and say so. Where they did not, a person at FP places the review on the scale, and the sentence they placed it on is printed beside it. That placement is the part you are entitled to distrust, which is exactly why it is shown instead of summarised.
A review is thrown out only for a reason we can state and you can check.
There are three: it arrived as part of a coordinated wave; it is a verdict with no supporting sentence; or its author has a declared interest in the film. Nothing else is a reason. Disliking what a review concludes is not a reason, and the negative reviews in these readings are counted at full weight.
The number is the mean of what was counted. That is the whole formula.
No weighting by outlet size, no confidence adjustment, no secret sauce. If a reading is thin, we print how thin rather than dressing the number up.
Readings are done by hand right now, and that is a limitation, not a feature.
The engine that reads sentences at scale does not exist yet. Until it does, a person does the reading, which means these numbers carry a person's judgment. The rubric above will not change when the engine arrives. Only the reader will.
We say where every quote came from.
Some reviews here we read in full at the source. Others we quote as they appear in a public encyclopedia entry, because we have not yet licensed access to the original. Each review says which, on its own line.