Oval vs Heart Shaped Face: Which Ruler Your Number Came From

The Forehead Width Each Row Wants, in Millimetres and Inches

One check does most of the work in this question, so it is worth stating what it is in full. Forehead width is taken across the widest part of your forehead, usually a little above the eyebrows; cheekbone width is taken across the two widest points of the face, just under the outer corners of the eyes. Divide the first by the second and you get a number that has no units, because it is a width divided by a width. Above 1.00 means the forehead is the wider of the two; below it means the cheekbones are.

This site publishes 0.90 as the centre of that row for an oval face and 1.05 for a heart face. The centre is the middle value the site puts each outline at, and this row sits on the scale a tape measures in. Each outline also carries a spread: the rule-set page, which lists every number the site works from, prints it as 0.08, or 0.0809 before it is rounded.

The table below is those two centres multiplied out against three cheekbone widths, so the ratio becomes a number your tape can actually reach. The three widths are the ends of the adult cheekbone band this site publishes (128.2 to 148.8 millimetres) and its working line in the middle (136.0); where that band comes from is named on its round-face glasses page and is not re-derived here. Choosing three widths is a presentation decision, not a claim that faces come in three sizes — every cell is one multiplication, so put your own measured width in place of the printed one and the row is yours.

Read the last column before the difference. The two ±1σ bands this site publishes for this row (a band is the range of forehead widths the site still accepts around each centre) overlap: at the working line, the stretch of forehead width that is consistent with an oval centre and consistent with a heart centre at the same time is 1.5 millimetres. Two bands whose centres are 20.5 mm apart cannot help crossing when each half-width is 11.0 mm: each band reaches further than half the gap. That is arithmetic, not a rounding artefact, and the overlap is not a third label.

In the units you can hold up against your own face: if your forehead measures roughly 122 millimetres (4.8 in) against a 136 mm cheekbone, you are near the middle of the oval row. At 143 mm (5.6 in) you are near the middle of the heart row. And across the strip from 131.8 to 133.3 millimetres (5.19 to 5.25 in) at that cheekbone width, both rows still accept you. Nothing here decides which of the two words you should use. It tells you how far apart the two published centres are in the units you can check, and how narrow the shared ground is.

Millimetre cells are the unrounded centres (0.89926 and 1.05) times the row's cheekbone width, printed to one decimal; inch cells are that same unrounded product divided by 25.4 and rounded separately, so an inch can land a hundredth away from dividing the millimetres beside it. Multiplying the printed 0.90 instead of the unrounded centre gives 115.4, 122.4 and 133.8 mm — the first two cells differ by 0.1 mm, and that difference is rounding, not measurement. The spread column is one published standard deviation of this row (0.0809 of a ratio on the tape scale, printed as 0.08 on the rule-set page) at that width, so it is a spread inside one outline, not a tolerance on your tape. The shared-width column comes from the unrounded band edges, which is why it is 1.4 to 1.6 mm while the printed two-decimal edges give a shared strip 0.01 of a ratio wide. One standard deviation is a chosen band width, not a natural boundary: on the same three widths, two spreads push the oval band to 0.74–1.06 and the heart band to 0.89–1.21, and the ground both rows allow becomes 23.5 mm (0.93 in) at the working line — which is where every useful distinction in this table disappears, so the table stays on one band width and says so.

Two Rulers, One Row, and the Count That Moves

This site classifies faces from a photograph. The numbers that classification runs on are not the numbers a tape gives, and the two are connected by a single straight-line conversion: multiply, then add a fixed amount, applied to the centres and to the spread together. The rule-set page publishes the tape version of the table, because that is the version a reader can check at home. The consequence for this row is specific and it is the reason this page exists.

On the tape scale, one of the seven outlines clears 1.00 on the forehead row: heart, at 1.05. On the classifier's own numbers, none of the seven clears it. Heart is still the highest of the seven there — 0.978 — and oval sits at 0.937, so the ordering between these two rows survives the change of ruler untouched. What does not survive is the sentence. On the tape, a heart outline's forehead is wider than its cheekbones. On the classifier's numbers, no outline's forehead is wider than the widest point of the face, heart included.

There is a detail in that second column worth looking at, because it is not a rounding matter. Heart's ±1σ band on the classifier's scale runs 0.956 to 1.000, printed as 0.96 to 1.00: its upper edge lands exactly on the mark and does not cross it.

The reason is arithmetic, and it is visible in two constants published on the rule-set page — the centre is 0.978 and the spread on that row is 0.022, and 0.978 plus 0.022 is exactly 1. So the highest forehead any row reaches on that scale reaches the point of equality and stops there, while the same row on the tape scale sits 0.05 past it. Two published descriptions of one outline, and the yes-or-no answer to 'is the forehead the widest part' comes out differently.

What that means in plain terms: the phrase 'wider forehead' is not a test you can run across two sources. A page that prints 1.05 and a page that prints 0.978 can be describing the identical row on two different rulers, agreeing about which outline sits highest and disagreeing about whether it clears 1. Before you compare anybody's number — including this site's — ask which scale it is on. Comparing one scale's number against the other scale's number is how two people end up arguing about the same face.

Centres are printed here the way the site's renderer prints them: the tape column carries 0.89926 rounded to 0.90 and the classifier column 0.937 rounded to 0.94. On the classifier scale the two printed bands meet at 0.96, so the shared strip has no width at all on the printed edges and 0.003 unrounded; on the tape scale it is 0.01 printed and 0.011 unrounded. Both are the same fraction of a spread — 0.136 — which is not a coincidence: the conversion scales centre and spread by one factor, so anything measured in spreads of that spread is identical on both rulers. That is exactly why the gap column reads 1.86 on both lines while the 'rows clearing 1.00' column reads 1 and 0.

Why the Distance Does Not Move When the Ruler Does

Two divisions, one per scale. On the tape: the difference between 1.05 and the unrounded oval centre is 0.1507, and this row's published spread on that scale is 0.0809 — printed as 0.08 on the rule-set page — so 0.1507 divided by 0.0809 is 1.86. On the classifier's numbers: the difference between 0.978 and 0.937 is 0.041, and the spread there is 0.022, so 0.041 divided by 0.022 is 1.86. Same answer, and it is the same answer because the conversion is a straight line applied to centres and spread together — a wider ruler makes both numbers bigger by the same amount, so the number of spreads between two rows cannot change.

That is the whole finding, stated once. Switching rulers cannot widen the gap between an oval outline and a heart outline. What switching rulers does is change the shape of the check: on the tape the gap sits above a mark (1.00) that one row crosses and the other does not, so the check reads as a yes-or-no question about the forehead. On the classifier's numbers the mark sits above every row, so the same check reads as a distance from 1.00 — which outline comes closest, not which one gets over. One of those two framings is a fact about the rule set and the other is a fact about the units it is printed in, and they are easy to confuse because both appear in the same sentence.

Weight the result honestly, because this site publishes the weighting elsewhere. The oval row in its table comes from eighteen measured faces; the heart row does not come from a measured group at all — its centre was interpolated along the literature's ordering of outlines from a single labelled photograph in the reference set. So 1.86 is a distance between two published numbers, one of which is less settled than the other, and it is not a measurement of how far apart real faces are. This site prints that caveat on the heart guide and on the heart result page, and it is repeated here rather than hidden.

Strip away the notation and what is left is this: whoever tells you that a heart face has a wider forehead than an oval face is right about the ordering and says nothing about the units. The ordering is the durable part. The number you can check at home is the tape one.

What This Site Already Prints About This Row

Before adding anything, it is worth being straight about how much of this is already on the record here, because the honest answer is: most of the two numbers, and none of the sentence about them. The oval outline guide carries a card for this pair on its confusion list, and its first line names the forehead row as the check doing the work. The sentence after that is the plain description: on an oval the forehead is narrower than the cheekbones, on a heart it is wider, and the jaw tapers away below it, with the chin row pointing the same way.

The same guide also answers the specific situation of a photo result and a self-image disagreeing, by telling you to check the forehead ratio on your result page and adding that the tape-measure calculator is an independent second read on the same question.

The heart outline guide prints the tape-side claim in its own way: the forehead row is the only row in the whole table sitting above the cheekbone width, which is what makes this outline the one that runs wider at the top than anywhere below it. The brow guide for a heart face prints the classifier-side claim, and prints it precisely — 0.978, the highest of the seven, meaning the forehead is 97.8 per cent as wide as the cheekbones, the only outline here whose forehead approaches the widest point.

The diamond-versus-heart page prints the count: heart is the only one of the seven rows above 1.00, diamond sits at 0.80, and the two published bands do not touch. Further down the same page teaches a reading of the strip, telling you to treat two numbers within 0.09 of each other as inside the band.

So the two halves of the sentence quoted at the top are already published, in two places, each of them correct on its own scale, and neither of them labelled with the scale it belongs to. What is new here is the collision, and the count. One of those pages says an outline crosses the mark; the other says the highest outline approaches it. Both are about the identical row of the identical table, and the way you tell them apart is section two, not a re-measurement.

There is a third thing already on the record, and nothing below adds a fourth version of it: which rows the chin ends up in. The oval beard guide already names the chin row as the check that draws the oval-and-heart line, and the diamond-versus-heart page already works out how many cells two rows can print identically. Those pages own their arithmetic; here they get a pointer, nothing more.

Which Boundary the Site Names, and Which One It Reports

The site's own guide for this outline says something unusual out loud, and it belongs on a page about this pair, so it is quoted rather than summarised. On the heart guide, under the table of the five checks, the rule set names diamond as the nearest boundary for this shape, because that is the pair it has a written explanation for. Read by the same constants, an oval outline actually sits slightly closer to heart than diamond does.

Two consequences follow, and they run in opposite directions, which is why both need stating. The first is about reading a result. If your numbers are close to the heart row, the outline whose bar sits closest on the page may well be oval rather than the boundary named in the text. The guide's own advice there is to read both comparisons and not treat the named one as the verdict.

The second consequence is about the page you are on. The pair named in that sentence — oval against heart — is not one of the six pairs this site has registered confusions for. Its runtime table gives oval against oblong and against round, round against square, heart against diamond, diamond against oval, triangle against square. So the boundary callout never prints this pair, and when a photograph result and a tape result disagree about oval or heart, what you get is the margin — the distance between your numbers and each of the seven outlines — printed with the site's confidence lowered, which is its way of saying it is less sure, not a note written for this border.

That asymmetry is the reason a question like this one keeps coming back. The site has a written explanation for heart against diamond, and it prints that explanation in several places. It has no pair-specific text for heart against oval, even though the outline its own nearest-boundary sentence calls closer is the one people ask about. This page is that missing pair, with one limitation attached. It can print the row and the two scales, and it can say what the runtime does and does not report. What it cannot do is turn the ordering inside that confession into a number to quote here: the distance table that would carry it is a different page's subject, and its figures are not moved onto this one.

In plain terms: nothing on your result page will tell you 'you are between oval and heart'. You have to read the row yourself, and section one is where it is written down in full.

The Second Check, and Why a Tape Cannot Give It

The order to run these checks in is fixed: the forehead row first, then the same row on the other scale, and the chin only if the first two come out level. So the chin row deserves one paragraph, and it deserves it with its limitation printed beside it rather than underneath.

Chin taper is the width at the very bottom of the face divided by the width of the jaw, and it is the row whose two ends are what people mean when they say a soft chin against a pointed one. On the tape scale this site publishes 0.45 for an oval row with a band of 0.37 to 0.53, and 0.30 for heart with a band of 0.22 to 0.38. That is 1.79 spreads apart, nearly as much as the forehead row's 1.86, and a wider margin than any of the three remaining checks. The two bands still share a strip 0.01 of a ratio wide, exactly as the forehead row's do.

Here is the limitation, and it is not a small one for anyone reading this with a tape: that row is the one a soft tape cannot supply. The chin's narrowest point is ambiguous by enough millimetres to move the whole band, which is why the calculator route this site publishes has no field for it and lets the photograph path carry that check alone. That fact is already stated, in full, on the diamond-versus-heart page and on the inverted-triangle guide. Nothing is added to it here; only the consequence is repeated. If you are working from your own measurements, the row carrying the second-largest margin on this pair is the row you cannot measure, and the two checks you can are the ones in the sections above.

So at home you have one usable check with a real margin behind it, not two. That is a property of tape measures, not of your face.

The Other Three Rows, Where the Two Almost Coincide

Run the same subtraction across the rest of the row and the picture sharpens by getting duller. Length ratio: oval 1.45 against heart 1.35, which is 0.54 of a spread, with published bands 1.27–1.63 and 1.17–1.53 sharing everything between 1.27 and 1.53. Jaw ratio: 0.80 against 0.72, 0.69 of a spread, bands 0.68–0.92 and 0.60–0.84 sharing 0.16. Jaw angle: 122° against 120°, 0.27 of a spread — the smallest gap of the five — with the two printed bands overlapping by 14 degrees out of a 16-degree band.

So on three of five checks these two outlines are closer than one spread of their own variation, and the shared strips are wide rather than narrow. Two rows carry more than a spread between them: the forehead row and the chin row — which is precisely the row whose yes-or-no answer changes with the ruler and the row a tape cannot give. That is a more useful description of this border than a list of adjectives about curves and points, because it tells you why the same face can be talked into either label. Three of the five readings you would bring to the question barely move between the two outlines, and the two that do move are the two with the awkwardness attached.

This site's classifier does not decide on one row anyway — it scores all five readings together and reports the margin against each outline. Nothing above should be read as 'the forehead row gives the verdict'. What it gives is the row where the two outlines are furthest apart in a scale you can measure yourself, which is a different and much weaker claim, and the weakest honest one available here.

What the Pages That Answer This Question Actually Print

Searching heart vs oval face shape leads to a set of pages that mostly answer with descriptions, and the descriptions are consistent: an oval outline is called balanced with a rounded chin, a heart outline is called wide at the forehead and narrowing to a pointed chin. One of them puts the pair in a single cell, and its wording is worth quoting because it is the closest any reachable page comes to a test: oval faces have similar widths at forehead and jaw, with a rounded chin; heart faces have wider foreheads and narrower, pointed chins. Similar to what, and wider by how much, are not answered anywhere on that page.

Numbers do appear, in two places, and neither lands on this border. One analysis site gives a centimetre example for a heart face — forehead 15, cheekbones 13.5, jaw 11, length 19 — alongside the instruction that forehead width exceeds cheekbone and jaw measurements. A men's styling guide gives each of six outlines a row of four inch numbers, with heart's forehead largest and the oval row's description reading length against cheekbones rather than forehead against cheekbones.

Take those two printed examples at face value and something odd shows up: that guide's oval example has 8.5 inches of forehead against 7.5 of cheekbone (8.5 ÷ 7.5 = 1.13) while its heart example has 8.5 against 8 (8.5 ÷ 8 = 1.06). Both of its outlines clear 1.00, the oval one by the larger margin, and its oval description never says the forehead is narrower. That is not a trap set for the reader; it is what happens when a page gives worked examples without stating the scale the examples are on.

A third page does state a division with a number on it: forehead at least 1.15 times the jaw, offered as the rule for a heart face. That is a genuine threshold and it is not usable for this pair, because its denominator is the jaw. The row this site and the two pages above are arguing about is forehead over cheekbones. A threshold on a different denominator cannot settle it, which is the reason the date-sorted table below records that row as a number on the forehead rather than a number on this border.

And there is a block above all of those, generated for this question rather than written by anyone. It describes a heart face as having a wider forehead or cheekbones that taper to a small pointed chin, and gives one number in its entire comparison: a length about one and a half times the width. That 'or' is worth naming, because it is where this whole question goes wrong: which of the two widths is wider is the check, and writing the pair of them as alternatives removes the only testable part of the sentence. The generated block is a rewrite of the pages beneath it and is not counted as a position anywhere above.

Here is what is actually on offer: nobody you can reach is giving you a forehead-to-cheekbone comparison for this pair. Two are giving you widths, one is giving you a threshold against the jaw, and the block at the head of the results has turned the relationship into a choice between two body parts.

Every Dated Position on This Border, Oldest First

The rows below are every position found for this question that carries a date, oldest first, including this site's own. Identities are given by type rather than name for every row that is not this site. Nothing outside this site's rows puts a number on the forehead-over-cheekbone comparison for this pair, and nothing outside this site's rows publishes a band.

Identities are given by type, not by name or link, for every row that is not this site, and no row here is offered as a comparison of this site against anybody else. Dates are the publication date printed on the page with the revision date where the page carries one; four of those come from machine-readable metadata on the page itself, and one row has no date printed on it at all, which is stated in its date cell rather than quietly used. Three rows share the publication date 2026-09-12 — this site's oval guide, heart guide and heart result page — and no order inside that date is claimed; they are listed by the date each was last revised. The 1901 row carries a year rather than a printed day, so it sits at the head of 1901 and no order inside that year is claimed. Not in this table, and not evidence for any sentence above: two pages behind a bot wall that returned an interstitial instead of an article, one image page with no article behind it, one discussion page that carries only a relative timestamp, and one short video with a caption and no readable body. The generated answer block at the head of the results for this question is a rewrite of the pages above it, gives one number in its entire comparison, and is not counted as a source. The 1901 row is quoted because it is the oldest dated statement found on this outline, not because age makes it right; on the one point in the title, it says the opposite of what this site's published row says.

What Moved, and What Did Not

Two spans sit in that table and they should not be run together. The dated web rows run from 2024-09-20 to 2026-07-14: one year and nine months, five positions, and the format has not moved an inch. Every one of them describes the outline rather than measuring the border. Four of the five name the forehead as the difference; two give absolute widths; one gives a threshold against the jaw; none gives the forehead-over-cheekbone comparison for this pair, with a band or without one.

The other span is 1901 to 2026-07-14: one hundred and twenty-five years. That longer one is where the genuinely interesting thing is, and it is not a story of progress. The beauty manual places the greatest width of an oval face between the temples, level with the outside corners of the eyebrows — above the cheekbones. This site's oval row puts the widest point at the cheekbones and the forehead at 0.90 of it, and the men's guide quoted above puts more forehead width on its oval example than on its heart example. A century and a quarter of vocabulary, three different answers to which part is widest on an oval face, and no reader is told which convention a given page is writing in.

What did move over the shorter span is the willingness to print a number at all, and it moved in the wrong place. The 2026 pages are more numerate than the 2024 ones: centimetre examples, inch examples, one explicit threshold with a percentage attached. Not one of those numbers is on the comparison the title asks about, so the additional precision buys nothing. Meanwhile the honest hedges got sharper — the newest external page in the set warns its reader to allow half an inch to an inch of margin on its own measurements, which is a stated tolerance, and none of them states a scale.

So the gap here is not that the forehead row went unnoticed. It is noticed on every page above, this site's included. It is that the row is stated as a relation, a set of examples and a threshold against three or four different denominators, and the one thing nobody prints beside it is which units it is in. That is a small-sounding omission with the specific consequence laid out in section two. The sentence 'a heart face has a wider forehead' is true here and false there, and a reader holding a photograph-derived number and a tape-derived number has no way to find out they are comparing two different rulers. The fix on any of those pages is one clause long: name the instrument the number came from.

The Order to Run It In

Four measurements you take yourself, on the scale section one is written in: the calculator prints your forehead row against all seven outlines and shows the margin, rather than naming one boundary and stopping. Face shape calculator ↓

If you want a usable answer rather than another description, run it in this order, and the first step is the one nobody lists. One: find out which scale your number is on. A soft tape gives you the tape scale, and any number taken from a photograph analyser is on that tool's own scale — which is not this site's scale, not necessarily, and not comparable to a tape number even when both are labelled 'forehead ratio'. A number in a screenshot with no stated scale cannot be checked against anything, including this page. Two: only then compare, and compare inside one scale. Three: if the row comes out level on both rulers, only then bring in the chin — knowing that if you are working from a tape, you cannot supply it.

For step one, the question to ask a page is not 'what is your number' but 'what is your number divided by, and in what units'. This site's answer is published: the rule-set page prints the tape-scale table, the same table the detector's numbers map onto by one straight-line conversion, and the result page beside your photo carries the converted version so it can sit next to your own measurements. If you would rather run the check with your own numbers than with a photograph's, the calculator below takes four measurements: length, forehead, cheekbones and jaw. The forehead over cheekbones row in it is on the tape scale, so section one's table applies to it cell for cell.

For step two, take the whole comparison rather than a rule of thumb. The two rows sit 1.86 spreads apart on either scale; the ground they share is 0.01 of a ratio wide on the tape and none at all on the printed edges of the classifier's scale; and at the working cheekbone width the two centres are 20.5 millimetres apart. Put your own width in the first table and those become your numbers.

For heart shaped face vs oval decisions made in public — in a comment on someone's face, in a quiz result you are about to act on — the useful question is the one asked from the first section onward: which ruler produced that answer. Two people can hold the same published row, one reading 1.05 and one reading 0.978, and disagree about whether the forehead is the widest part of a face without either of them being wrong about the number they are holding.

What This Page Does Not Prove

It is worth finishing with what a reader should not carry away. Nothing on this page establishes that an oval outline or a heart outline is more attractive, more common or more versatile. This site publishes no prevalence figure for any of the seven shapes, for the reason it prints on the outline guides: the figures in circulation have no stated sample and no stated method, so there is nothing to check them against. The heart shaped vs oval face question, asked in either order, is a question about measurements, and it stays one.

It does not prove that the two published rows are far apart. They are 1.86 spreads apart on one row and less than one spread apart on three of the five; the classifier scores all five readings together; and the outline this site's own nearest-boundary sentence calls closer than the named boundary is this pair. A reader whose numbers sit in the shared strip is not doing arithmetic wrong.

It does not settle the 1901 row against the 2026 rows, and it deliberately does not try. The oldest statement in the date table puts the widest point of an oval face above the cheekbones. This site's published row puts it at the cheekbones. The difference is which convention a page writes in, and a page one hundred and twenty-five years older than the rule set is evidence about vocabulary, not evidence about a face.

And the arithmetic is only as good as its weakest input, which this site states elsewhere and repeats here. The heart row was interpolated from one labelled photograph rather than fitted across a measured group, so every number here that involves that row inherits it: the 0.05 that clears 1.00, the 20.5 millimetres, the 1.86 spreads.

Widening the band from one spread to two, as section one's footnote shows, turns a 1.5 millimetre shared strip into a 23.5 millimetre one, which is enough to swallow the practical difference between the two centres. Any of the arithmetic here can be redone with your own width and your own two readings in place of the published ones. If it stops being useful when you do, your face is the reason, and the arithmetic is not arguing with it.

Questions this page answers