Some axes exist because users expect them

Not every property people want scored is something a photograph can support. "Confidence," "erotic impact," "how it feels in person" are all things a rubric designer might be asked to include, and none of them sit cleanly in pixels the way lighting or contour do. Rubrics include axes like this anyway, and the reason is less about the model's capability and more about what a rubric looks like without them.

The gap between complete-looking and observable

A rubric with a hole in it - no axis at all for something users care about - reads as incomplete, even when the missing property genuinely cannot be scored from a photo with any reliability. Product teams feel this pressure directly: a six-axis rubric that skips the one thing every user actually wants to know about looks like it is dodging the question, whether or not dodging is the technically honest move. The result is that some axes get included for completeness even though the model behind them is doing something closer to a guess dressed as a score than a read.

What "unmeasurable" means here

This is not the same claim as what a photo geometrically cannot contain - the absence of scale information from a 2D projection, which is a hard physical fact true of every model regardless of training. An unmeasurable-in-practice axis is softer: the model can output a number for it, the number has some correlation with the labels it was trained on, but the underlying property is subjective, contextual, or dependent on information a single photo simply does not carry. "Erotic impact" is a genuine case of the second kind - it is not blocked by geometry the way size is, but it is downstream of taste, context and a viewer's own state in a way no rubric definition fully pins down, which makes it structurally harder to score consistently than something like surface condition, however precisely the axis gets operationalised on paper.

How a designer should label an axis like this

The honest move is not to drop the axis, and it is not to present it identically to a more mechanical one either. It is to be explicit, inside the product, that some axes are closer to inference and taste than others - the same distinction multi-axis scoring versus one number draws between axes that are "measurement-adjacent" and axes that are frankly subjective. A rubric that names this difference is doing its users a service; one that presents every axis with identical visual weight and identical confidence is quietly overstating what the softer axes actually deliver, which is the same caution penisrater.com gives users reading a total that folds a taste axis in with the rest.

Why this differs from a simply hard subject axis

A subject axis being hard to read is often a matter of degree - a model can be less reliable on shape than on lighting while still being reliably better than chance, and reliability improves somewhat with better training data. An axis in this category is different in kind: more labelled data narrows disagreement among labellers only slightly, because the underlying property is genuinely contested even among people looking at the same photo with full context, not merely hard for a model to infer from a partial one. No amount of retraining converts a taste axis into a measurement axis; better data makes the model's guess more consistent with the taste of the people who labelled it, which is not the same achievement.

Reading a rubric that includes one

Rate Cock scores an axis in this territory alongside more mechanical ones, and treating every axis on any rubric with the same confidence is the mistake worth avoiding, whichever tool you are reading a chart from. A human judge handles subjective axes more honestly by default, since a person on ratepenis.com can simply say "this is a matter of taste" in the review itself rather than folding taste into a number that looks as precise as a measured one. The physical axes at the other end of this spectrum - the ones a tape measure actually settles - are covered by measuremycock.com, and the contrast between the two is the clearest way to see where a rubric's softer axes actually sit.