How we test
Written so that anyone could repeat it. If a result on this site cannot be reproduced from this page, the result is wrong, not the page.
Two of every shirt, at retail, with our own money
One goes through the wash. The other stays sealed in its bag as the control, so every color reading is a comparison against the same shirt from the same batch, not against a manufacturer swatch. We do not accept samples. Brands do not know they are being tested until the results are up.
Weigh, measure and photograph before the first wash
Mass in grams on a kitchen scale, converted to gsm from the measured area. Chest width at 2.5 cm below the armpit and body length from the high point of the shoulder, both in inches, both on a flat shirt. Then the cycle-zero photograph, which is the reference every later frame is compared to.
Thirty cycles, all shirts in the same load
Warm water, a standard supermarket detergent at the dose on the bottle, tumble dry on low, shirts inside out. This is a deliberate choice. Care labels say cold and line dry; almost nobody does that, and a shirt that only survives ideal treatment is not a shirt we would recommend.
Photograph at washes 0, 5, 10, 20 and 30
Flat, same camera, same spot in the same room, same fixed light, from the same distance, with a color calibration card in every single frame. Phone cameras auto-correct white balance and exposure, so a raw photo would measure the room, not the shirt. In the editor, the white balance eyedropper goes on the card's neutral gray square first, which pulls the whole frame back to true color. Only then is the shirt sampled, always at the same point: 10 cm below the center of the collar seam, on the chest. The hex code that comes out is the swatch you see on the Gray Line page.
Report dE2000 against the sealed control
Color difference is reported as dE2000, the current standard formula, against the unwashed control photographed under the same setup. We call the first wash where a shirt crosses dE 3.0 its Gray Line. For white shirts the same setup measures yellowing as a shift in CIE b*, and opacity as the visibility of a printed target behind a single layer of fabric.
Every number, including the ones that hurt our pick
The raw table goes up for download when a run completes. Nothing is smoothed or dropped. If the shirt we recommend fades first, that is the result, and the recommendation changes.
What this test is not
It is not a laboratory test. AATCC 61 and AATCC 135, the industry standards for colorfastness to laundering and for shrinkage, use controlled machines, specified detergents and calibrated instruments we do not have. Our method is modeled on them and is repeatable, but a dE reading from a corrected photograph is not a spectrophotometer reading. Treat the numbers as a fair comparison between shirts tested identically, not as absolute values you could compare against a lab report.
It is also not finished. The first run started in September 2026 and thirty washes take real time. Pages say "in testing" wherever a reading does not exist yet, and we would rather show you an empty cell than a number we made up.