The face rating scale runs 1 to 10, but it does not behave the way people assume. Across 17,032 analyses run between January and August 2026, the mean score was 5.59 and the median was 5.3. The midpoint of the scale is 5, and the midpoint of the actual population is 5.3, which sounds like a small difference and is not, because the distribution around it is extremely tight.
Half of everyone, 8,450 people out of 17,032, scores between 5.0 and 5.9. Add the band below and 74.4% of the dataset is inside the two-point strip from 4.0 to 5.9. The scale has ten points and most people are living on two of them.
The full distribution, band by band
This is the whole scale with real counts attached.
| Score band | People | Share | What it means |
|---|---|---|---|
| Below 4.0 | 77 | 0.5% | Very rare, and rarer than people expect |
| 4.0 to 4.9 | 4,230 | 24.8% | Below average, and the second largest band |
| 5.0 to 5.9 | 8,450 | 49.6% | The modal result, half of everyone |
| 6.0 to 6.9 | 2,891 | 17.0% | Top quarter |
| 7.0 to 7.9 | 824 | 4.8% | Roughly 1 person in 20 |
| 8.0 to 8.9 | 524 | 3.1% | Top few percent |
| 9.0 and above | 36 | 0.2% | Statistical outlier |
The shape is a narrow peak with thin tails on both sides. That second point matters as much as the first: 0.5% of people scored below 4.0 and 0.2% scored 9 or above, so the bottom of the scale is almost as empty as the top. Anyone whose private expectation of themselves is a 3 is, on this evidence, describing a band that 77 people out of 17,032 occupy.
What each threshold is worth as a percentile
The cumulative view is the more useful one, because a score only means something relative to the people it is being compared against.
| Score | Share scoring at or above |
|---|---|
| 5 | 74.7% |
| 6 | 25.1% |
| 7 | 8.1% |
| 8 | 3.3% |
| 9 | 0.2% |
The percentiles from the other direction say the same thing. The 90th percentile of all scores is 6.8, the 95th is 7.8 and the 99th is 8.8. So clearing 7.0 puts you past roughly 92 people in 100, and clearing 8.8 puts you in the top 1%.
Why a 7 feels worse than it is
This is the single biggest misreading of the scale, and it comes from school. Twelve years of 70% being a C teaches you that 7 out of 10 is unremarkable. On this distribution, 7.0 is the 92nd percentile. It is not a C. It is a result that 91.9% of people do not reach.
The gap between the felt meaning and the measured meaning is enormous at every level. A 5.3 feels like failure and is the exact median. A 6.0 feels average and is the top quarter. A 7.0 feels mediocre and is roughly 1 person in 12. An 8.0 feels like the minimum for "attractive" and is reached by 3.3% of people. We have written the individual bands up separately: is a 4 out of 10 a good face rating, is a 6 out of 10 a good face rating and is an 8 out of 10 a good face rating each go through one band in detail.
The correct habit is to translate every score into a percentile before reacting to it. The number is arbitrary. The percentile is the information.
Why the middle is so crowded
The crowding is not a quirk of this particular model. It is what happens when you measure a population against a prototype.
A rating is built from twelve separately scored features, and those features are themselves tightly clustered: they average between 5.97 for eyes and 5.30 for skin, a range of just 0.67 points. Combining twelve mid-range measurements produces a mid-range result for almost everybody, and the only way to escape the middle is to be high on all twelve at once rather than spectacular on one.
The data on high scorers shows exactly that. People scoring 8 or above average 8.90 on bone structure, 8.89 on jawline, 8.66 on facial harmony and 8.54 on symmetry, with no feature below 7.66. There is no weak link. Meanwhile, across everyone, the average gap between a person's best and worst feature is only 1.38 points, and 58.0% of people tie for their weakest. Most faces are flat, which means most scores are middling.
What "objective" does and does not mean here
People search for an objective face rating, and it is worth being precise about which part of that is achievable.
The score is objective in the sense that it is repeatable and calibrated. Re-uploading the same photo to the same account normally returns your stored score, the model runs at temperature 0 so repeat scores are stable, the scoring is anchored to a distribution of 17,032 real results rather than to a vibe, and the model has no incentive to flatter. Symmetry, the strongest single component, correlates with the final score at 0.95, and even golden-ratio compliance, which correlates at only 0.67, is the model’s estimate of a geometric property rather than an opinion about taste.
What it is not is a verdict on attractiveness. It scores facial geometry and presentation from one still image. It cannot see expression in motion, voice, style, humour or presence, and those carry enormous weight in how anyone actually experiences a face. A calibrated number about geometry is a useful data point and a bad summary of a person.
There is also a sampling caveat that applies to every figure on this page. People who run a face rating tool are not a random sample of the population, so "top 8.1%" describes the people who showed up, not humanity.
How to place your own score
Three steps, and none of them involve comparing yourself to a screenshot.
First, get a clean reading. Front-on, at eye level, in even light, neutral expression, hair off the face. Symmetry is the most photography-sensitive component of the whole score, and a five-degree head turn is genuinely read as asymmetry.
Second, find your band in the first table and your percentile in the second. That is the entire interpretation. A 5.6 is slightly above the median and inside the largest band in the dataset. A 6.4 is in the top quarter. A 7.2 is roughly 1 in 12.
Third, look at the feature breakdown rather than the headline. A single number tells you where you sit; the twelve components tell you why, and which of them are worth attention. The attractiveness test returns the score with your percentile against this distribution attached, so the placing is done for you rather than left to guesswork, and the complete band-by-band figures are in the average face rating study.