Research & methodology
Every ratio Kouros measures and every estimate it gives is traced back to where it actually comes from: a published standard, established research, or an honestly-labeled in-house heuristic.
Kouros never claims to measure your body fat percentage the way a lab instrument does. Instead, it gives you a trained visual estimate. That's the same kind of read an experienced coach or bodybuilder gives at a glance, based on cues like muscle striations, vascularity, skin thickness and tightness, how much definition shows versus how smooth the skin looks, and where any softness sits across your frame.
Because that read can't honestly be pinned to one decimal point, Kouros always gives you a range instead of a single fake-precise number. A clear, well-lit, unambiguous photo typically gets a tight range, around 2-4 percentage points wide. A photo with worse lighting, an awkward pose, or looser clothing gets a wider range, because the estimate is genuinely less certain, not because of a fixed formula.
This matters because body composition is inherently hard to pin down precisely, even with real clinical tools (see the comparison below). Visual assessment by an experienced eye is a practical, well-correlated method for exactly that reason, and that's what this estimate is: a visual read, not a diagnostic measurement.
None of these methods are as precise as they get credit for, including the clinical ones. Here's how they actually compare.
Clinical gold standard
The most accurate widely available method, but still carries roughly 3-4 percentage points of error and varies by machine calibration and hydration. Requires a specialized clinic visit and isn't something most people can get regularly.
Consumer scales, handheld devices
Convenient and cheap, but considerably less reliable than DEXA, and highly sensitive to hydration and technique: the same person can read several points differently across a single day depending on when they last ate, drank, or trained.
Gyms, trainers
Can be accurate in trained hands, but the result depends heavily on which sites are pinched, how firmly, and by whom, making at-home self-measurement especially inconsistent between sessions.
What this app does
A trained AI visual read of the same cues an experienced coach uses for an “eye test”: muscle striations, vascularity, skin thickness and tightness, fullness versus definition, and how fat is distributed across the frame. Presented as an honest range, not a fake-precise single number.
Further reading: DEXA / DXA, bioelectrical impedance analysis, and body fat percentage measurement methods on Wikipedia.
Every ratio Kouros scores, with its real evidentiary basis. Where a genuine published standard exists, we say so and link it. Where none exists, we say that too, plainly, rather than dressing a heuristic up as an established norm.
The direction (bigger shoulder-to-waist taper reads as more attractive) has genuine research behind it. The specific 1.618 number is a popular fitness-marketing convention, not a figure derived from a study.
Waist-to-shoulder ratio (WSR) attractiveness research
Horvath T. (1981). "Physical attractiveness: the influence of selected torso parameters." Archives of Sexual Behavior, 10(1), 21-24. · Braun MF, Bryan A. (2006). "Female waist-to-hip and male waist-to-shoulder ratios as determinants of romantic partner desirability." Journal of Social and Personal Relationships, 23(5), 805-819.
Real, peer-reviewed evolutionary-psychology research consistently finds a tapered male upper torso (narrower waist relative to shoulders) rated more attractive, in study-reported ranges broadly consistent with this app's 1.618 target. But the specific 1.618 figure itself is not drawn from these studies: it's inherited from the commercial "Adonis Index" / "Adonis Golden Ratio" fitness-marketing program (Kyle Leon, c. 2011), which in turn borrows the golden-ratio-in-classical-art tradition (Pacioli's Divina Proportione, 1509). That broader tradition is itself disputed by historians and mathematicians as applied to art, architecture, and anatomy. Claims that the golden ratio governs the Parthenon or human body proportions are widely regarded as unproven (see Livio, M., The Golden Ratio, 2002; Wikipedia's "Golden ratio" article, "Disputed observations" section, verified live during this research pass). Treat 1.618 as a memorable, culturally popular target that happens to sit inside the range real attractiveness research supports, not as a number computed from that research.
This is the one metric in the whole catalog backed by a real, numbered, health-body-issued threshold, tying waist taper to abdominal-obesity risk rather than pure aesthetics.
WHO abdominal-obesity threshold (waist-to-hip ratio)
World Health Organization. (2008). Waist Circumference and Waist–Hip Ratio: Report of a WHO Expert Consultation, Geneva, 8–11 December 2008. WHO defines abdominal obesity as WHR > 0.90 (men) and > 0.85 (women).
Directly verified this cycle: the 0.90/0.85 cutoffs are real and match this app's 0-0.9 band exactly. Caveat worth keeping in mind: the WHO protocol measures true circumference with a tape at defined anatomical landmarks under a clinical procedure. This app derives WHR from 2D photo-landmark width, a proxy, not a literal circumference, so the threshold is directionally correct, not a clinically precise equivalent.
Shoulder-to-hip ratio is a real, independently studied attractiveness cue, not just a derived combination of other ratios, but the exact 1.3 target used here hasn't been confirmed against the study's reported number.
Shoulder-to-hip ratio (SHR) attractiveness research
Pazhoohi F, Arantes J, Kingstone A, Pinal D. (2023). "Neural Correlates and Perceived Attractiveness of Male and Female Shoulder-to-Hip Ratio in Men and Women: An EEG Study." Archives of Sexual Behavior, 52(5), 2123-2141.
Shoulder-to-hip ratio is studied in its own right in the attractiveness-research literature, not merely as an arithmetic combination of shoulder-to-waist and waist-to-hip, which upgrades this from the existing source string's framing ("heuristic, tunable, derived from combining..."). That said, I could not extract the specific preferred numeric SHR value the cited study reports (I confirmed the paper is real and on-topic via its title/journal/DOI, but did not obtain full-text access this cycle), so this app's 1.3 target itself remains unverified against that number and should still be treated as an untraced internal heuristic pending a full-text check.
Real research backs the idea that a wider chest relative to waist reads as more attractive; the specific 1.4 number is a bodybuilding-culture convention layered on top of that finding, not lifted directly from the studies.
Waist-to-chest ratio (WCR) attractiveness research
Fan J, Dai W, Liu F, Wu J. (2005). "Visual perception of male body attractiveness." Proceedings of the Royal Society B: Biological Sciences, 272(1560), 219-226. · Swami V, Tovée MJ. (2008). "The Muscular Male: A Comparison of the Physical Attractiveness Preferences of Gay and Heterosexual Men." International Journal of Men's Health, 7(1), 59-71.
Fan et al. is a frequently cited finding that waist-to-chest ratio was the strongest single predictor of male body attractiveness ratings in their sample, ahead of BMI and waist-to-hip ratio: real, on-topic, peer-reviewed support for chest-to-waist as a meaningful ratio to track. As with shoulderToWaistRatio, the specific 1.4 target (±0.1) is inherited from informal bodybuilding-culture convention (commonly cited as ~1.3-1.5:1 across fitness articles and forums) rather than a number computed from these papers, which I could not verify.
A real, decades-old physical-therapy convention for spotting forward-lean posture, adapted here into a percentage score rather than the clinical absolute-distance measurement.
Kendall plumb-line postural assessment
Kendall FP, McCreary EK, Provance PG, Rodgers MM, Romani WA. Muscles: Testing and Function with Posture and Pain (5th ed.). Lippincott Williams & Wilkins, 2005 (book's existence and prior editions confirmed via Open Library this cycle; first edition by Kendall & Kendall, 1949).
Real, widely used physical-therapy textbook convention: in ideal standing alignment, a vertical plumb line should pass through the ear lobe, the acromion (shoulder), the greater trochanter (hip), slightly anterior to the knee axis, and slightly anterior to the lateral malleolus (ankle). This app measures ear deviation from the shoulder-hip line as a percent of torso height, rather than the textbook's absolute-distance-from-plumb-line protocol, and the ±6% tolerance is this app's own choice, not lifted from the book.
This is a real clinical cutoff, and the app's own 48-degree threshold matches the literature's cutoff almost exactly.
Clinical craniovertebral angle (CVA) / forward-head-posture cutoff
Representative literature confirming the ~48° cutoff (C7-to-tragus line vs. horizontal) for forward-head-posture classification: e.g. a paper in the International Journal of Health Sciences and Research, Vol. 12, Issue 6 (June 2022) states "48 degrees is defined as Forward head posture." Broader CVA methodology reviewed in Mahmoud NF, Hassan KA, Abdelmajeed SF, Moustafa IM, Silva AG. (2019). "The Relationship Between Forward Head Posture and Neck Pain: a Systematic Review and Meta-Analysis." Current Reviews in Musculoskeletal Medicine, 12(4), 562-577.
Verified live this cycle via search: the ~48-50° cutoff for forward head posture is a real, widely cited clinical convention, matching this app's min:48 threshold closely. CVA itself (also called "cervical angle") is an established, frequently studied clinical measure, confirmed via multiple current papers (Physiopedia's Craniovertebral angle page, MDPI, Taylor & Francis, and others) discussing CVA reference values and cutoffs.
This is the same real, decades-old plumb-line convention as posturalAlignment, just read at a different one of its five reference points.
Kendall plumb-line postural assessment (same convention as posturalAlignment)
Kendall FP, McCreary EK, Provance PG, Rodgers MM, Romani WA. Muscles: Testing and Function with Posture and Pain (5th ed.). Lippincott Williams & Wilkins, 2005.
Worth correcting a small overstatement in the existing informal `source` string, which frames the knee point as this app "extending" Kendall's convention down to the knee. The knee is already one of the five canonical points in Kendall's own plumb line (ear lobe, shoulder, hip, slightly anterior to the knee axis, slightly anterior to the ankle). So this isn't a novel extension, it's directly using another point from the same textbook convention already cited for posturalAlignment. The ±6% tolerance and hip-to-ankle-length normalization are this app's own choices.
Backed by the same real WHO threshold as the front waist-to-hip ratio, but read from an angle the WHO protocol doesn't actually define; treat it as a cross-check, not a second independent data point.
WHO abdominal-obesity threshold (waist-to-hip ratio), same source as the front waistToHipRatio
World Health Organization. (2008). Waist Circumference and Waist–Hip Ratio: Report of a WHO Expert Consultation, Geneva, 8–11 December 2008.
Same underlying WHO threshold as waistToHipRatio, applied from behind. The back view is an even less standard vantage for WHR than the already-approximate front-view proxy (WHO's protocol doesn't specify a rear-view method at all), so this metric's result is best read as a secondary corroboration of the front-view read (useful when the front photo's pose makes that measurement less reliable), not an independent clinical measurement in its own right.
The underlying idea (symmetry reads as more attractive) is real and researched in general, but this exact formula and its tolerance were designed in-house, not measured in any study.
Body-symmetry research broadly links bilateral symmetry to attractiveness (e.g. Wade TJ. (2010). "The Relationships between Symmetry and Attractiveness and Mating Relevant Decisions and Behavior: A Review." Symmetry, 2(2), 1081-1098), but no study defines this app's specific construction: left/right waist deviation from the neck-to-hip midline, as a percent of waist span. This is a geometric formalization of the general "symmetry matters" finding, not a value drawn from a study; the 0 target and 4% tolerance are this app's own empirical choices.
This isn't really measuring an aesthetic ideal; it's flagging when the photo itself was taken at an angle or pose that skews the other measurements, using navel drift as the tell.
No standard exists, and this metric is different in kind from the others: it mostly detects camera/pose artifacts (a torso lean or twist during the photo) rather than a body-shape trait anyone rates for attractiveness. It's closer to a photo-quality/pose-consistency check dressed as a physique metric than an aesthetic ideal, so 'heuristic' undersells how little it has to do with an ideal proportion in the first place. The 0 target and 5% tolerance are this app's own empirical choices.
Research supports that bigger, more developed arms generally read as more attractive, but no study ties that to this specific arm-width-to-waist-width ratio or its 1.4 target.
No study establishes a specific arm-circumference-to-waist-width numeric ratio, unlike the chest/shoulder cases above where WCR/WSR research exists. General muscularity-attractiveness research does support the broad direction: Frederick DA, Haselton MG. (2007). "Why Is Muscularity Sexy? Tests of the Fitness Indicator Hypothesis." Personality and Social Psychology Bulletin, 33(8), 1167-1183, found perceived muscularity (including arm size) tracks attractiveness ratings, but that's a study about overall muscularity as a fitness signal, not a specific arm:waist proportion. The 1.4 target and 0.15 tolerance remain an untraced internal heuristic.
The reasoning behind this metric (lats drive back-width perception) is sound, but no study has actually measured attractiveness preferences for back-view lat-to-waist proportions the way front-view chest/shoulder ratios have been studied.
Unlike the front-view shoulder/chest ratios, I could not find any published research evaluating back-view or lat-spread-specific attractiveness ratios: the WCR/WSR psychology literature above is uniformly front-facing-photo based. The existing source string frames this as "the back-view analog" of the front shoulder-to-waist ratio, and that anatomical reasoning (lats, not shoulders, drive perceived back width) is sound, but it is an extrapolation of front-view research to a view that research doesn't actually cover, not an independently verified standard. The 1.618 target is a direct copy of the front ratio's target, not a number derived for this view.
Purely an in-house design choice about how much trap width should show without competing with the V-taper shape.
No literature found relating trapezius width to lat width as a proportion. The 0.75 target reflects an internal design goal (well-developed traps should add visible upper-back thickness without visually crowding out the V-taper the lats create) rather than a citable source.
Same evidentiary basis as the front waist-symmetry metric: symmetry-matters is a real general finding, but this specific formula and tolerance were designed in-house.
Same situation as waistSymmetry: general body-symmetry-attractiveness research exists but no study defines this exact geometric construction (left/right waist deviation from the spine midline as a percent of waist span). Geometric formalization of a real general finding, not a value drawn from a study. The 0 target and 4% tolerance are this app's own empirical choices.
No specific research on this ratio exists. The target used to assume this measured a single calf's thickness, but it actually measures the span between both calves' outer edges, so the number was reset to match what a normal reference photo actually shows for that span, not a measured attractiveness standard.
No literature found relating calf circumference/width to waist width as an attractiveness ratio. General muscularity-attractiveness research (Frederick & Haselton, 2007, cited under armToWaistRatio) supports that well-developed muscle groups generally read as more fit/attractive, but says nothing about this specific proportion. Corrected this cycle (ORDER-140): the previous 0.5 target/0.08 tolerance was calibrated as if this measured a single calf's own width, but `calfL`/`calfR` are each placed at the *outer* (lateral, away-from-midline) edge of their respective calf, mirroring `latWidthL`/`latWidthR`'s across-body span design, so the geometry actually measures the span between the two calves' outer edges (closer to a stance-width read than an individual muscle-thickness read), which any normal standing photo will show as much wider than half the waist width. Re-derived the target from the hand-verified reference back photo (`scripts/physique-landmark-example-coords.json`'s `calfL`/`calfR` vs `waistBackL`/`waistBackR`, cross-checked against the reopened `calfL.jpg`/`calfR.jpg` crops): the raw span ratio there measures ~1.22. Tolerance widened to 0.2 (looser than the torso-anatomy ratios) because stance width naturally varies more photo-to-photo than a fixed anatomical span, and only one hand-verified reference photo backs this number.