Oblong and Oval Face Shape: Which of the Four Borders Is Yours?

Oval or oblong face shape: the measurement the borders are drawn on

The measurement is a quotient of two lengths, which is why it carries no unit. Face length runs from the centre of the hairline to the tip of the chin. Divide it by the width across the tops of the cheekbones, point to point, and you have the ratio that every one of these borders is stated in. Two of the three guides divide by the cheekbone width. The third divides by the average of three widths — forehead, cheekbones and jaw — which on most faces is a smaller number than the cheekbone width, and therefore produces a larger ratio from exactly the same face.

That is not a rounding difference. Take the worked example that guide printed: a face 22 cm long, with widths of 14, 15 and 13 cm. The average of the three widths is 14, and 22 ÷ 14 = 1.57, which the guide that printed it calls oblong. Divide the same length by the cheekbone width of 15 instead and you get 22 ÷ 15 = 1.47, which falls inside that same guide's own borderline band of 1.45 to 1.5 — the one it names the long oval. Nobody measured anything wrongly, and nobody misread a table. The two numbers came from one face and one tape, and they landed on opposite sides of a border.

Hold that for everything below: a ratio without its denominator is not a measurement, it is half of one.

Slightly oblong face shape: what the borderline band covers

“Slightly oblong” is not a category any of the three publishes. The nearest thing in print is the gap one of them opened between its oval and its oblong — 1.45 to 1.5, which it calls a long oval and advises styling as an oval under one borrowed oblong rule. The band on this site is wider, and it is published rather than buried: the rule set centres oval at 1.45 on the tape scale with a tolerance of ±0.18, and oblong at 1.65 with the same tolerance, so the two ranges share everything between 1.47 and 1.63.

Those two centres need a caveat, because the honest version is less tidy than the number. Neither 1.45 nor 1.65 is an average of measured faces. Both come from a calibration: the ordering of the seven shapes in the published literature was mapped onto the range this site's detector actually measures, endpoint to endpoint, which is why oblong lands exactly on the top of that range and oval 0.20 below it. The ±0.18 is a pooled within-group standard deviation taken from a set of 50 photographs whose draft labels were applied by one person. /about prints the whole of it, including the two things that calibration cannot prove.

The consequence is the point. Two bands that overlap across 0.16 of a range 0.36 wide are not two boxes a face drops into. They are two conventions whose shared ground covers most of the range a real face can land in, and a face sitting in that shared ground is being labelled by choice of convention rather than by measurement. This is also why the result page names a second shape: when the runner-up reaches 0.60 of the leading probability, the result says so instead of hiding it, and on that 50-photo calibration set the hint fires for 32% of faces.

Inside the band the ratio is the wrong instrument, so use the side line and the three widths. On an oval the widths step down — cheekbones widest, forehead next, jaw narrowest — and on the tape scale this site publishes that as a jaw ratio of about 0.80 for oval against about 0.92 for oblong, jaw width over cheekbone width. Those two bands touch rather than separate, overlapping from 0.80 to 0.92, but the direction holds and it is the direction you can see in a mirror: the closer your jaw width sits to your cheekbone width, the straighter your side reads.

Centimetres in, inches out: the same face, the same border

A centimetre tape and an inch tape cannot disagree about your shape, and it is worth seeing why, because some of the confusion in this family is unit confusion wearing a different hat. One inch has been exactly 25.4 mm since 1959, so converting is exact arithmetic and not an approximation.

Worked, on the 22 cm face above. 22 cm ÷ 2.54 = 8.66 in. The three widths: 14 cm = 5.51 in, 15 cm = 5.91 in, 13 cm = 5.12 in. Divide in centimetres and you get 22 ÷ 14 = 1.57. Divide in inches and you get 8.66 ÷ 5.51 = 1.57. The quotient is identical to two decimals, because the ratio is one length over another length in the same unit and the unit cancels. Converting never moves you across a border. Changing the denominator does.

Six readings, both units, one quotient each:

Every inch figure is the centimetre figure divided by 2.54, exact since 1959. The 1.50 row is the whole argument in one line: a face measured to the millimetre, called oblong by one published border and oval by two others, with no error anywhere in the arithmetic.

One tape reading, five lines

The other direction, on a face measured in inches. Someone holding an inch tape reported a face 6 to 7 inches long and about 5 inches wide, and asked which shape that was. In centimetres that is 15.24 to 17.78 cm long and 12.70 cm wide. The ratio is 1.20 at the short end and 1.40 at the long end, in either unit, and it is worth running that one pair of numbers against everything currently published.

One face, five verdicts: oval twice with a caveat at the bottom of the range, twice below both shapes entirely, and once a reading that only makes sense on a scale nobody prints. The answer given at the time — possibly oval, but it is hard to tell from measurements alone — was the only one of the five that described the situation rather than guessing at it.

The last row is not a border but a scale, and it is the reason the other four disagree with your tape. Across the 50 photographs in the calibration set the detector's own length ratio spans 1.075 to 1.305, because its landmarks sit below the hairline and outside the cheekbone arch. It is not the number a tape gives, which is why every figure this site shows you is converted to the tape scale before it is printed.

The four borders, in the order they were published

Every position on this border that could be dated to a day, oldest first — 30 June 2013 to 12 September 2026, thirteen years and two months. The third column is the one to read down.

Identities are given by type, not by name or link, for every row that is not this site. Dates are the publication date printed on the page or, where a page carries none, the date the position was written down.

Take the four borders with you. The full chart — the three measurement lines, all four borders on one number line from 1.20 to 1.85 with the overlap shaded, the denominator switch worked both ways, a centimetre-and-inch strip, and a two-method log to fill in with the tape in your hand — is a file you can print or keep on your phone: Download the boundary chart (printable) ↓

What moved in thirteen years, and what did not

Read down the third column and one thing is obvious: the number moved four times in the ten months between November 2025 and September 2026, and it moved in both directions — 1.6, then 1.5, then 1.75, then 1.65. Read down the second column and the other thing is obvious. The positions that offered no number at all, in 2013 and again in 2026, offered the same criterion on both occasions, and it was the side of the face. Thirteen years and two months of answers, and the only thing that has been revised is the arithmetic. The geometry has not moved once.

The one other number in the table disagrees as well. The November 2025 guide puts an oblong jaw angle at 105 – 115°; this site publishes 118° with a tolerance of ±8°, which is 110° to 126°. The two ranges share five degrees. It is the same story on a smaller axis: two conventions, drawn for two different purposes, overlapping where a real face is likely to sit.

Mixed face shape, and why the overlap is not a defect

Some faces sit in the shared band and stay there however carefully you measure. That is a mixed result in the only sense the words support: two conventions each describe part of the outline and neither describes all of it. It is not an eighth shape, and it is not a failure to measure properly.

The evidence that this is ordinary rather than rare is not statistical, because nobody has the statistics. It is the split verdicts. Ten people looking at one photograph produced six ovals, three hearts and one round-oval mix. Three people looking at another produced oblong-with-diamond, oval and heart inside a single day. One printed guide photographed a face for its oval section and the same face for its diamond section, and the reader who caught it had found both images in one search.

What a mixed result should change is which advice you read, not which label you claim. Every recommendation attached to oblong in this family exists to break a vertical line: weight at the sides, a fringe covering part of the forehead, a brow-heavy frame. Each of those is keyed to length, not to the word. If your ratio sits in the overlap and your side line reads straight, the length recommendations apply and the oval ones do not, whatever you end up calling the shape. If your side line curves, the reverse holds. The label is the shorthand; the dimension is the thing.

One claim in the table deserves a specific answer, because it keeps recurring: that the rules run differently for men and for women. Nothing in the four published borders is sexed — all four print one set of numbers for everybody. What does differ is how the same outline reads, since a soft jawline looks more defined on a face carrying more masseter, and an oval outline with the jaw widened lands at square rather than at oblong. That is a difference in what the geometry looks like, not in where the border sits.

What a photograph of your face cannot see

Every ratio on this page came either from a tape or from a photograph, and the two are not measuring the same object. A camera records a projection, and that projection depends on how far the lens stood from the face and what focal length it used: the same face at different distances yields different proportions in the image. The point was made in 2013 with two demonstrations attached, and nothing about lenses has changed since.

This site's detector reads landmarks rather than tape points, and the landmarks are not the points a tape touches, so its raw quotients sit systematically away from tape readings — across the calibration set the length ratio came out about 0.18 lower, the jaw ratio about 0.09 lower and the jaw angle about 13° higher, with the forehead ratio close to equal. That 0.18 is not the ±0.18 tolerance quoted earlier on this page, and the two agreeing is a coincidence worth flagging: one is a standard deviation carried across from the detector scale by the mapping's slope, the other is a fixed distance between two sets of measurement points. A tolerance tells you how wide a band is; an offset tells you where the band sits. Confusing them is how a face ends up two bands away from itself.

Neither path is the truthful one. A tape in your own hands has a different failure mode: you cannot see the point you are measuring to, so the centre of the hairline and the apex of the cheekbone get placed by feel, and two millimetres on a 22 cm length moves 1.57 to 1.59 — two in the second decimal of a ratio whose published borders are stated to two decimals.

The order to run it in

Five steps, and the order is the argument: the side line first, the ratio last, and the ratio only ever against one named border.

• Look at the side first. Hair pulled back, mirror at eye level: does the outline from temple to jaw corner drop straight, or does it curve in? Straight is oblong, curving is oval. This is the only criterion in thirteen years of answers that nobody has revised.
• Then take the three widths — forehead at its widest, cheekbones across the tops, jaw corner to corner — and see whether they step down or run level. Divide jaw by cheekbones: near 0.80 steps down, near 0.92 runs level.
• Then take the length, centre of hairline to chin tip, and divide by the cheekbone width. Write the denominator down beside the number, in the same breath.
• Compare against one border, named: 1.5, 1.6, 1.75, or this site's 1.65. If you compare against all four you are not measuring, you are collecting verdicts.
• If the number lands inside the overlap, stop using it as a tiebreaker and style against the dimension instead. In this family, length is the only dimension either label is about.

Questions this page answers