Cut registry › Jewellery cuts
Jewellery cuts · 21 shapes, one protocol
What each shape costs you, in light.
Every popular jewellery cut, put through the same ray census as the historic and house cuts in the registry, and compared with the one benchmark we did not get to choose: Tolkowsky’s 1919 round brilliant.
The short answer
No fancy shape beats a round brilliant for light return, and none of the 21 here comes close. That is not a failure of cutting. A round brilliant is the only outline that can be optimised for light alone; every other shape begins with a decision about how it should look and fits the optics around it.
What differs is the price of that decision. The cushion gives up 4.92 points of Global against the standard. The baguette gives up 32.05. Between those two sits every shape most people actually choose between.
Each figure below measures our reference geometry for that shape, on one protocol, so the shapes are comparable with each other. None of them is a grade for an individual stone.
The board
Sorted by Global, the weighted score used across this site: 0.34 useful light, 0.20 fire, 0.16 tilt, 0.10 scintillation, 0.10 symmetry, 0.10 low leak. The standard is T57, in the tinted row.
| Cut | Facets | Useful light | Leak | Fire | Global | vs standard |
|---|---|---|---|---|---|---|
| Square Step cut | 69 | 85.08% | 9.66% | 72.45 | 84.11 | −4.49 |
| Cushion Fancy brilliant | 85 | 79.83% | 9.31% | 73.28 | 83.68 | −4.92 |
| Trillion Fancy brilliant | 73 | 66.74% | 12.65% | 87.32 | 80.86 | −7.75 |
| Trilliant Fancy brilliant | 70 | 78.04% | 11.05% | 83.37 | 80.48 | −8.12 |
| Octagon Step cut | 49 | 71.24% | 17.59% | 76.47 | 79.91 | −8.69 |
| Marquise Elongated brilliant | 87 | 64.06% | 17.23% | 88.52 | 79.73 | −8.88 |
| Asscher Step cut | 58 | 69.28% | 11.33% | 80.17 | 78.36 | −10.25 |
| Square radiant Square brilliant | 49 | 72.42% | 12.37% | 62.12 | 76.82 | −11.78 |
| Square emerald Step cut | 57 | 63.72% | 17.32% | 83.95 | 76.47 | −12.14 |
| Princess Square brilliant | 69 | 56.92% | 18.08% | 82.99 | 76.20 | −12.40 |
| Hexagon Step cut | 37 | 58.18% | 24.23% | 84.83 | 75.93 | −12.68 |
| Oval Elongated brilliant | 93 | 46.85% | 26.99% | 92.46 | 74.96 | −13.65 |
| Pear Elongated brilliant | 64 | 53.36% | 21.60% | 83.80 | 71.55 | −17.06 |
| Teardrop Elongated brilliant | 74 | 46.02% | 28.13% | 89.08 | 71.40 | −17.21 |
| Triangle Step cut | 19 | 52.81% | 25.42% | 86.38 | 71.23 | −17.37 |
| Emerald Step cut | 57 | 55.35% | 25.50% | 87.11 | 70.48 | −18.13 |
| Rectangle Step cut | 45 | 50.20% | 32.32% | 77.08 | 69.23 | −19.38 |
| Radiant Square brilliant | 60 | 31.55% | 41.49% | 76.55 | 65.14 | −23.46 |
| Heart Fancy brilliant | 62 | 46.65% | 32.47% | 83.80 | 62.85 | −25.75 |
| Pixel Brilliani house design | 32 | 36.54% | 50.73% | 78.96 | 62.74 | −25.86 |
| Baguette Step cut | 25 | 40.43% | 38.00% | 62.11 | 56.56 | −32.05 |
| T57 the standard | 57 | 92.28% | 1.88% | 78.85 | 88.61 | 0.00 |
A measurement nobody else publishes
Light return is usually quoted as a single number for a whole stone. That hides something useful. Binning every returning ray by the point on the crown it leaves through, in five equal bands across the stone, shows how evenly a shape spreads what it returns.
| Cut | Centre bands | End bands | Difference |
|---|---|---|---|
| Oval | 83.1% | 41.9% | 41.2 |
| Emerald | 79.6% | 37.6% | 42.1 |
| Pear | 77.8% | 44.3% | 33.5 |
| Marquise | 86.8% | 57.5% | 29.3 |
| Princess | 66.2% | 40.7% | 25.5 |
| Cushion | 88.2% | 85.7% | 2.5 |
| T57 control | 99.1% | 98.1% | 1.0 |
A round brilliant is flat across its whole face, which is what a working control should read. An emerald cut and an oval lose more than forty points between their middles and their ends. The princess loses it at the corners, which are also the corners that chip. Same protocol as every other figure here, measured in diamond.
All 21
Questions
Of the eleven jewellery cuts measured here, the cushion scores highest overall in diamond at 83.68 Global, followed by the trillion at 80.86 and the marquise at 79.73. All of them sit below the round brilliant standard, which measures 88.61. No fancy shape beats a round brilliant for light return, and any page claiming otherwise is selling something.
Because a round brilliant is the only outline that can be optimised for light return alone. Every other shape starts from a decision about how it should look and fits the optics around it. The gap ranges from 4.92 points for the cushion to 32.05 for the baguette. That is the price of the shape, and it is worth knowing rather than hiding.
No. GIA issues cut grades for round brilliants only. Every fancy shape receives symmetry and polish grades, which describe finish rather than light performance. That absence is the reason proportion ranges carry so much weight in fancy-shape buying advice, and the reason we measure the shapes ourselves.
A bow-tie is the dark band across the waist of an elongated brilliant such as an oval, pear or marquise. It is a shadow rather than a leak: the facets beneath it draw their light from the part of the sky the viewer's own head and shoulders block. We tested our reference geometry for a bow-tie under both face-up and full-hemisphere illumination with a modelled observer and found none, so we publish the mechanism and not a severity figure. We would rather show an empty column than a number we cannot stand behind.
The cushion. It returns 79.83% of entering light usefully against the round brilliant standard's 92.28%, for a Global gap of 4.92 points, the smallest on this page. Its light is also the most evenly distributed of the fancy shapes.
No. Each figure measures our reference geometry for that shape at the proportions published on its page, not an average of stones on the market. Treat them as a like-for-like comparison between shapes on one protocol, not as a grade for any individual stone. A cut to different proportions will read differently, which is the whole reason proportions matter.
T57, Marcel Tolkowsky's 1919 ideal round brilliant: 57 facets, a 53% table, a 34.5 degree crown and a 40.75 degree pavilion. We did not choose those numbers, which is exactly why we use them. It measures 88.61 Global in diamond on this protocol.
Elongated and shallow shapes spread furthest. On our depth-derived spread axis the asscher and trillion lead the group, both above 106. That axis reads millimetres per carat from depth and does not model outline area, so for outline-driven shapes such as the marquise and oval you should compare face-up millimetres rather than rely on it.
Method
Monte-Carlo on exact facet geometry: 3,200 rays face-up plus 2,200 at 20° tilt, cosine-weighted from a 7° near-vertical cone. Useful light is what exits through the crown within 64° of vertical. Identical protocol on every cut on this site.
Our reference geometry for each shape at the proportions published on its own page, not an average of market stones. A like-for-like comparison between shapes, not a grade for any individual diamond.
The Index symmetry axis scores facet regularity within each pavilion family. That is informative for a round brilliant and misleading for shapes whose pavilions are deliberately irregular, so we leave it out of these pages entirely.
Engine estimates, ±2 points, not lab-certified grades. GIA issues no cut grade for any of these shapes, so nothing here should be read as one.
The house and historic cuts live in the cut registry. Every shape here can be rendered and turned in the Index or set against another in Compare.