Scores

Axis, defined

An axis is one component the tool scores separately before combining; the word promises independence it does not always deliver.

By Updated 2 min readScores

Guides on Scores: How much weight a rating tool result deserves, Every part of a rating result, and what each one is for, What can and cannot be compared, and how

An axis is one component a tool scores on its own, before combining it with the others into a total. A tool that reports proportion, symmetry, and presentation separately is reporting three axes, and the word implies each was judged independently of the other two.

That independence is the whole point of having axes rather than one number, and it is also the part a tool can quietly fail to deliver.

One example

A tool scores a submission 7 on proportion, 6 on symmetry, 8 on presentation, and totals it to 7. If those three numbers were computed by genuinely separate passes, changing only the framing of a photo should move presentation while leaving the other two roughly where they were. If instead all three came from one internal judgement split three ways for the results page, all three will move together no matter which single thing changed - the axis labels are real, the independence behind them is not.

The caveat that matters most

The way to tell the difference from the outside is to watch what happens across a few submissions. When every axis moves together across your results, you have one axis wearing several labels, not several judgements. Genuine axes drift apart sometimes - one up, one down, one flat - because they are responding to different things in the image. Axes that always rise and fall as a set are telling you the breakdown is decorative, whatever the results page implies. Human scorers fail the same way, and it has a name: assessing resident physicians, Thomas and colleagues (2011, Journal of General Internal Medicine) describe the halo effect as "a good or bad performance in one area affects assessments in other performance domains", and measured an inter-item correlation of 0.68 across individual faculty ratings.

This is worth checking before trusting a breakdown for anything more specific than a total, including attributing a change in your score to a specific thing you changed. It is a smaller, sharper version of the same question a rubric answers at the level of the whole tool: whether the number in front of you was actually built the way its labels suggest.

An axis, in this sense, has nothing to do with a tape measurement along a physical dimension - that kind of axis belongs to measurement, not to a rating tool - and nothing to do with the register a human reviewer writes in, which ratepenis.com covers on its own terms. It is also downstream of whatever the model computes first: how that computation happens is one layer below where axes get assigned. Rate Cock is one tool built around six such axes with a public chart per entry, which is one way - not the only way - to make the independence claim checkable rather than asserted, because the six numbers sit side by side instead of folding into one figure.

Read next

Full archive