What the 1 to 10 scale is grounded in

The 1 to 10 scale on this page is the academic-style face-rating distribution familiar from facial-attractiveness research papers, where rater panels score photographs on a 1-to-10 (or sometimes 1-to-7) Likert scale and inter-rater agreement is measured statistically. The most cited synthesis is Langlois, Kalakanis, Rubenstein, Larson, Hallam, and Smoot (2000), a meta-analysis of 919 effect sizes confirming that attractiveness ratings are reliable across raters within and across cultures — establishing that attractiveness is not "all subjective" in the way the colloquial defence assumes.

The site's 1 to 10 band is mapped from the Face Score, a validated percentile from 1 to 99, not from the 17 landmark-based structural metrics reported alongside it. Their reference bands come from published anthropometry, including Farkas's normative measurements (Farkas 1994). Each metric score places your measurement against the first scans of about 2,100 RealSmile users, so 50 is the median scanner.

The bands below describe each tier in terms of the structural patterns that typically produce that rating in a rater-panel study: which metrics tend to be near the population mean, which tend to deviate in attractive-coded directions, and which tend to be flagged as drag metrics. This is the academic / clinical framing of the 1 to 10 scale. If you want the looksmax-community version of the same percentile distribution (sub5, LTN, MTN, HTN, Chadlite, Chad, Gigachad), the PSL chart page covers that translation explicitly.

The reference table

The figure beside each band is the Face Score range it comes from. The mapping is coarse: 39 or below maps to roughly 1–4, 40–59 to 5–6, 60–79 to 7–8, and 80+ to 9–10, so each range below 80 covers two bands.

1 to 2 — Below typical baseline

Face Score 39 or below

Structural markers: Significant facial asymmetry (visible at 2 meters), severe acne or scarring, structural anomaly that disrupts the 17-metric baseline, or any combination of three or more flagged metrics below the 5th percentile.

Highest-leverage move: Soft-tissue and grooming levers are the place to start. Structural concerns warrant a board-certified consultation before any DIY work.

3 to 4 — Below average

Face Score 39 or below

Structural markers: Two or three weak areas, most commonly skin texture, default expression and posture, none of which the 17 metrics read. The structural baseline is fine; the soft-tissue and expression layers are under-presented.

Highest-leverage move: This is the band with the largest available gains, and the levers are soft-tissue: consistent skincare plus posture and expression practice, no structural intervention. Re-check under the same photo setup each time, so a change in the photo is not mistaken for a change in the face.

5 — Average

Face Score 40–59

Structural markers: No metrics strongly flagged in either direction. Baseline skin, baseline expression, baseline structure. The face reads as neither distinctive nor disharmonious.

Highest-leverage move: Moving from 5 to 6 is typically a body composition lever (lower visceral fat reveals jawline) or a current haircut and grooming refresh. Moving from 5 to 7 means working several of those levers at once, re-checked under the same photo setup each time.

6 — Above average

Face Score 40–59

Structural markers: One or two metrics in the 75th percentile or higher, often an eye-region or jaw metric. The rest at baseline. The face reads positively in good lighting and average angles.

Highest-leverage move: Moving from 6 to 7 typically means finding and removing one specific drag. Often that is skin texture, default posture or under-styled hair, which none of the 17 metrics read, so check those by eye. On the measured side, the paid report names your two weakest metrics. From there it is targeted work, not generalized advice.

7 — Strong

Face Score 60–79

Structural markers: Multiple metrics in the 75th to 90th percentile range, often including Canthal Tilt, Jaw Taper and Facial Thirds Balance. Clear skin and a habitual Duchenne smile tend to come with a 7, though none of the 17 metrics reads either. The face reads positively across lighting conditions and angles.

Highest-leverage move: Moving from 7 to 8 typically requires structural assets that are largely set by the late twenties. Body composition, photography skill, and expression refinement carry the remaining gains.

8 — Very strong

Face Score 60–79

Structural markers: Multiple metrics in the 90th percentile or higher. Strong structural baseline (Jaw Taper, FWHR and Canthal Tilt all favorable), excellent soft-tissue (clear skin, controlled expression), and habitual photo competence.

Highest-leverage move: Maintenance band. Sleep, skin, body composition, and recovery sustain this band. Drift downward is more likely from neglect than upward from effort. The improvement plan in this band is a maintenance protocol, not a gain plan.

9 to 10 — Exceptional

Face Score 80+

Structural markers: Structural assets across the board, consistent across multiple photos and lighting conditions, plus expression and presentation that compound the structural baseline. Often appears in casting, modeling, or specific genetic clusters.

Highest-leverage move: No improvement plan in the looksmaxxing sense. Maintenance protocols (sleep, skin, body composition) and photography skill carry the entire upside in this band. Structural intervention is almost always a net negative at this level.

How rater-panel methodology maps onto a single user

  1. Mimic the rater-panel photo conditions used in attractiveness-research studies: neutral expression, frontal pose, hair off the forehead, no glasses, even diffuse lighting at face height. Most published attractiveness research deliberately removes the variables that move ratings without changing the underlying face (makeup, accessories, hair styling, dramatic lighting) so you should too when running this kind of measurement on yourself.
  2. Take three independent captures (different sessions, similar conditions). The published face-rating literature treats single-photo ratings as noisy and recommends multi-photo aggregation; running the scan on three captures and taking the median replicates that practice for a sample of one.
  3. Treat your median Face Score as the estimate, then read off the 1-to-10 band. The Likert band is what would round-trip into a published attractiveness study; the 1-to-99 Face Score is the higher-resolution number you should track over time.
  4. For a second reading, repeat steps 1 and 2 a week later. Do not expect the two sessions to match within a point or two: roughly half of the photo-to-photo variation in the Face Score lives between sittings, so more shots in one session cannot remove it. The median across both sessions is a steadier read than either one alone.
  5. Compare against the published age-and-sex-banded distribution rather than against the absolute 1-to-10 number alone. The Langlois meta-analysis and successor literature consistently report that attractiveness ratings are interpretable in age-and-sex context; an "objective" 6/10 on a 22-year-old means something different from a 6/10 on a 52-year-old.

Construct limits the academic literature already acknowledges

  • A rater-panel attractiveness score measures perceived first-impression appeal in a controlled photograph, not "attractiveness" as a global construct. The construct-validity caveats in Langlois et al. (2000) and Hönekopp (2006) apply: the rating predicts certain first-impression outcomes well and other social outcomes poorly.
  • Symmetry, averageness, and dimorphism are the three structural drivers the literature consistently isolates (Rhodes 2006, Annual Review of Psychology). They are necessary but not sufficient — a face that scores high on all three but is photographed badly will rate lower than its structural baseline would predict.
  • Inter-rater agreement is high in aggregate but individual raters disagree non-trivially. A 6.8 mean rating across 30 raters can hide a 4-to-9 spread; the band is a population-level summary, not a guarantee any specific viewer will react that way.
  • Cross-cultural validity is high for facial-symmetry and averageness preferences and weaker for specific dimorphism preferences (Cunningham et al. 1995). The chart bands describe structural pattern; absolute preference weights drift modestly across cultural contexts.
  • The chart is descriptive of where the face sits in the published distribution; it is not a prescription for any cosmetic or behavioural change. The structural-only framing is deliberately separated from clinical or community recommendation.

PSL scale vs a validated percentile — what "PSL-calibrated" actually means

The PSL scale is the looksmax-community relabeling of the same 1-to-10 distribution this page describes — sub5, LTN, MTN, HTN, Chadlite, Chad are percentile bands with forum names (the PSL chart page maps them one-to-one). A wave of 2026 rating apps now markets "PSL-calibrated" scoring. Read that phrase precisely: it means the app's output is tuned to agree with forum vocabulary, not that it was checked against how real people rate real faces. Calibration to a naming convention is not validation.

A validated score means the number was tested against human raters and the agreement was published. RealSmile's Face Score is cross-validated at r ≈ 0.8 against 1,566 human-rated faces (SCUT-FBP5500, 60 raters per face) — and we publish the failure too: our own earlier 17-metric geometry composite correlated at r ≈ −0.33 and was demoted to a measurement map. No "PSL-calibrated" app we have reviewed publishes either number. If you want your PSL tier, take the validated percentile and read the band off the chart — same distribution, honest number first.

Face rating scale FAQ

Why use a 1 to 10 scale instead of a 0 to 100 scale?+
The 1 to 10 rating is a coarser read of your Face Score, a validated percentile from 1 to 99: a score of 39 or below lands at roughly 1–4, 40–59 at 5–6, 60–79 at 7–8, and 80+ at 9–10. Most users find the 1 to 10 bands easier to interpret because it maps onto how people actually talk about faces in everyday conversation. The 0 to 100 scale is the one each metric is scored on in the optional full report.
What does a 7 out of 10 actually look like?+
On RealSmile, a 7 comes from a Face Score of roughly 60 to 79, the range that maps to 7–8. The reference table above describes the 7 band as rater-panel studies see it: often a positive canthal tilt, a defined jawline, balanced facial thirds, clear skin, and a habitual Duchenne smile in photos. The face reads positively across lighting and angles, not just in flattering ones.
Is the scoring objective or subjective?+
The measurement layer is objective: 17 structural metrics calculated from a 68-landmark detection model. Every metric your photo supports is scored 0 to 100 in the full report. Each metric score places your measurement against the first scans of about 2,100 RealSmile users, so 50 is the median scanner. The 1 to 10 rating is not built from those metrics: it is mapped from your Face Score, a percentile cross-validated against 1,566 human-rated faces.
Can I really move from a 5 to a 7?+
The levers exist, but there is no reliable timeline for a band move, and one higher reading is not proof of one: lighting, angle and expression move the score too. The highest-leverage starting points are well-documented in dermatology and exercise-physiology research: consistent skincare with daily SPF for texture and tone, body-composition work for visible structural definition, and shifting your default photo expression toward a Duchenne smile (AU6 + AU12 — see Ekman and Friesen 1978). The paid report names your weakest metrics and ties its plan to them — but we do not invent a guaranteed point gain we cannot defend.
Are the score bands biased toward Western features?+
The bias to know about is in the training data. The 1 to 10 band is read from your Face Score, and the model behind it was fitted to human ratings of SCUT-FBP5500, a research benchmark of posed studio portraits of mostly young, mostly East Asian faces. It learned what those raters liked in those faces, so faces unlike them are the ones it has seen least. The metric scores are a separate layer and do not feed the band.
How accurate is the score?+
The Face Score behind the 1 to 10 band is cross-validated at r ≈ 0.8 against 1,566 human-rated faces from a research benchmark. On 46 real phone photos rated by our own users it reaches r ≈ 0.40. It also moves between photos: across all the photos one person takes on the same day, the gap between their highest and lowest Face Score is a median of 11 percentile points. Lighting, angle, and expression cause more variance than measurement noise. We recommend running a few photos (neutral light, smile, three-quarter angle) and taking the middle of the range as a steadier read than any single shot.
What is the relationship between the score and dating outcomes?+
Correlated but not deterministic. The score predicts first-impression perception in photos, which is a meaningful piece of dating-app outcomes specifically. Real-world dating outcomes depend on photo quality, app strategy, prompt and message quality, location, and a long list of other variables. A higher score makes the photo layer easier; it does not bypass everything else.
Is the test free?+
Yes. Your Face Score (a percentile) and your strongest metric are free. The optional $14.99 Looksmax Report scores every metric your photo supports (typically 15 of the 17), identifies your two weakest, and writes a personalized improvement plan.

Free score is the headline. Full report is the plan.

Find out which band you sit in and what to do next.

The $14.99 Looksmax Report identifies your band, names your two weakest metrics, and writes a 4-week improvement plan targeted at your specific weak metrics. One-time price — no subscription required.

Get your face rating now

Free, instant, private. Your Face Score, the percentile behind the 1 to 10 band, and your strongest metric.

17 metrics · 68 landmarks · Photos auto-deleted