How the rating works
Your rating is not an opinion about your face. It is a position — where thousands of individual choices by real people place your photo among every other rated photo. This page explains exactly how that number is made, what makes it sharper, and what it will never pretend to know.
One question, two photos
Everything starts with the same matchup: two photos side by side, one question — who is more attractive? — and one tap. Nobody is ever asked to score a face from 1 to 10. Direct scoring collapses the moment two raters mean different things by a “7”; a choice between two photos means the same thing no matter who makes it. That is the whole trick, and it is why the ratings here are comparable when star-rating systems are not.
Chess math, not face math
Each matchup is scored with the rating system used in competitive chess (the Glicko‑2 refinement of Elo). Beating a highly‑rated photo moves your rating more than beating a low one, an upset counts for more than an expected win, and every photo carries not just a rating but a confidence interval that narrows as evidence accumulates. No part of the pipeline measures your face: no symmetry score, no golden ratio, no “canthal tilt”. The algorithmic face analysis other attractiveness tests sell is exactly the thing this system exists to avoid.
The number out of 10 is a percentile
Your displayed rating is your percentile against every actively rated photo, shown out of 10 — the 95.3rd percentile is a 9.53, the 50th is a 5.0. That means the number always answers a concrete question: what share of rated photos does the crowd place below yours? It is honest about being relative. There is no absolute beauty scale hiding underneath, here or anywhere else.
One honest caveat about that denominator: it is people who chose to be rated, and that group likely skews more confident — and plausibly better-looking — than the general population. A 5 here is the middle of a self-selected room, not the middle of humanity; if anything, scores here run harsh relative to a random crowd. We would rather say that plainly than recalibrate the scale to flatter anyone.
Precision is earned, never bought
A rating built on 30 comparisons is accurate to roughly ±0.5; at 150 comparisons it tightens to about ±0.2. The display follows the evidence: one decimal place until your photo has been through 150 head-to-head comparisons, and the second decimal only after. Paying for a boost sends your photo into more matchups sooner — it makes the evidence arrive faster. It does not and cannot change what the evidence says. Money buys speed, never a better number.
Real photos, live camera only
Photos enter the pool through a live camera capture with a liveness check — no file uploads, no filters, no photo of a photo. What gets rated is what you look like now, which keeps the comparisons meaningful for everyone in the pool. Every photo also passes moderation before it is ever shown to a rater.
Votes are weighed, quietly
Not every tap counts equally. The system watches for behaviour that looks like a script or a speed-run — impossible rhythms, one-sided tapping — and quietly discounts those votes so they cannot move anyone’s rating. An honest rater with unusual taste is never punished: disagreeing with the crowd is a preference, not an error, and your own “preference profile” on the results page treats it as exactly that.
Who picked you
Group breakdowns are being built: raters will be able to share optional demographics, and your result will show how different groups voted — where your rating stands among women 25–34, say, versus men 18–24. The rule is already set: a group needs 20 votes before its number is shown at all, because below that the data is noise, and noise is not something worth showing you.
What this is, and is not
Ratings here are the aggregated subjective opinions of other users, measured carefully. They are not a measurement of you, not a verdict, and not a prediction of anything that matters in your life. The system is built to be honest about uncertainty — coarse numbers until the evidence exists, no invented precision, no algorithm pretending objectivity about beauty. If you want a number made of real people’s actual choices instead of an AI’s geometry homework, that is the one thing this does.