Sharpness as a variable

Yes, blur moves a score: a slightly soft photo loses fine detail that some scoring axes read, so it can rate differently from a sharp photo of the same subject. Hand shake, or focus that landed just in front of or behind the subject, removes that detail before a rating tool ever sees the file.

Two kinds of blur, one effect

Focus blur happens when the camera's autofocus locks onto the wrong plane - close, if there is anything nearer in frame, or the background, if the subject lacks contrast for the sensor to grab onto. Motion blur happens during the exposure itself, from hand shake or subject movement, and gets worse in low light because the shutter has to stay open longer to gather enough light. Both blur types do the same thing to a rated image: they discard the fine texture and edge detail that some scoring axes are reading, replacing it with a smoother, lower-information version of the same subject.

A tool cannot tell a soft photo from a genuinely different subject. Image models are measurably sensitive to exactly this: Dodge and Karam (2016) tested classifiers against blur, noise, contrast and JPEG compression and found them "susceptible to these quality distortions, particularly to blur and noise." It can only score the pixels it is given, and blurred pixels carry less information than sharp ones - not a different subject, a worse copy of the same one, closer to what heavy compression or a screenshot does to the same file. The modelling underneath a rating tool reads whatever detail survives into the file, and blur removes some of that detail before the model gets a chance to weigh in on it at all.

Why this is a protocol issue, not a skill issue

The fix is not steadier hands. It is removing hand shake from the equation entirely: enough light to keep shutter speed short, a moment where the subject is still, and a firm brace or support rather than a free hand. None of that requires equipment beyond what most people already own, and it is the same logic a fixed support and a timer apply at the next level up, without needing that full setup here.

What to record

Sharp or soft is not usually a deliberate choice, which is exactly why it belongs in the same log as distance, light and crop - if a result looks off and the photo was slightly soft, that is worth ruling out before assuming anything changed about the subject. The same principle that makes a tape measurement independent of camera behaviour does not apply to a photograph, where equipment and handling are baked into every result whether you tracked them or not. A tool that exposes a breakdown, the way Rate Cock does on public entries, gives a better chance of noticing that one texture-dependent axis moved on a soft shot while others held - a total number gives no such signal, and a person looking at the photo directly would simply notice the blur and discount it, which a fixed rubric will not do on its own.