Screen stats: the first numbers from real screens

Every screen test on this site measures something real: the refresh rate your browser actually gets, the resolution and pixel ratio the screen reports, how many fingers the touch grid saw at once. For years those numbers lived and died in one browser tab. Since 2.23 they also feed a public page, screen stats, which counts what real screens measure, in groups big enough that no single device can be picked out.

The first groups crossed the publishing line this week. Below is what they say so far, and what changed in 2.24.

What changed in 2.24

We launched the stats with a bar across the top of the homepage and every test page: a checkbox, three links and a status line. It sat exactly where people come to run a test, and for anonymous hardware measurements it was more ceremony than the data deserved. In 2.24 it is gone.

• Nothing above the tools. Measurements join the stats quietly, as before. No new fields are collected.
• One off switch, on the privacy page. Tick “Don’t include results from this browser” on the privacy page and that browser stops sending screen stats. The tests work exactly the same either way.
• Hourly counts. The page used to wait for a nightly count that skipped the current day, so a result could take two days to show. It now refreshes every hour.
• Easier to find. Screen stats is linked from the footer on every language edition and from the site menu. The optional question after the dead pixel test and the privacy section are now translated.

Most screens are still 1080p-class, and most desktops still run at 1x

As of this morning the stats hold 508 screen readouts from the homepage and the resolution checker. The size class is the long edge of the physical resolution, in ranges, never the exact figure.

The 1,920 to 2,559 class covers both a 1080p monitor and most mid-range phones, which hold about 2,400 pixels on their long edge. One desktop in eight reports a 4K-class panel.

Pixel ratio says more about how people use those panels. On desktops, 48% run at 1x, 19% at 1.25x and 17% at 1.5x. Those last two are almost always Windows display scaling on a laptop or a dense monitor. Only 10% of desktops report 2x, the ratio a Mac Retina screen gives. On phones the common ratio is 3x, at 41%.

Phones are well ahead on color

Wide color and HDR are where phones and desktops part ways.

Nearly half the phones say they can show HDR. Fewer than one desktop in ten does, and that undercounts monitors with an HDR badge on the box: on Windows, browsers generally report HDR only when it is switched on in display settings. If you want to know what your own screen reports and what the terms mean, sRGB, P3 and HDR explained walks through it.

Refresh rate: 60 Hz is still the common answer

137 runs of the refresh rate test are in. Half of all runs, and 62% of phone runs, measured 60 Hz. On desktops the spread is wider: 38% at 60 Hz, 14% at 120, 13% at 240 and 10% at 144.

Two things in there are worth a closer look. First, 12% of phone runs measured 30 Hz. That is almost never the panel. Battery saver modes, Low Power Mode on an iPhone among them, halve the frame rate a browser gets. If your phone reads 30 Hz, turn the saver off and run it again. Second, some of the phones that read 60 can do 120. Many phones drop the rate when they decide nothing on screen needs it, and some browsers are kept at 60 unless you change a setting. The test measures what the browser actually receives, which is the honest number, but it is not always the number on the spec sheet.

Touch: one finger, and the ghosts

335 touch screen test grids are counted. On phones, 72% only ever saw one finger at a time. That does not mean those screens track one finger. It means most people paint the grid with a single fingertip, which is the right way to find a dead zone. Two fingers is the next most common answer, at 19%. Treat this number as a record of how people test, not of what the hardware can do.

The ghost touch test is the one we were most curious about. Of 158 hands-off runs, 78% came back clean, 4% caught one or two stray touches, and 18% caught three or more, which is what the test calls ghost touch. On battery alone it was 20%. That is far more than the share of phones with a real fault, and it should be: people run a ghost touch test because something already feels wrong. The charging group is still under 30 runs, so the comparison between plugged-in and battery phones has to wait.

Touch sampling rate, from 49 phone runs so far, lands most often at 120 Hz (33%) and 125 Hz (22%). That is the rate the browser hands touch events to the page, which on many phones is capped near the display rate, not the faster rate the touch panel samples internally.

Dead pixel reports need you to say what you found, so they come in slowly: 12 so far. They appear on the page at 30.

How the counting works

Each sample carries the test it came from, the device type (phone, tablet or desktop), the OS family and browser family without a version, the date to the day, and the measurement itself. No IP address is stored, no identifier, no user agent, no exact resolution, no country and no language. Each browser sends at most one sample per kind of measurement per day, so one person rerunning a test does not tip a chart. A group is shown only at 30 samples, and rare values under 5 are folded into “Other”. The privacy page lists every field.

These are screens tested here, not a random sample of all screens in the world. That leans every number toward people who had a reason to test, and toward the phones and monitors our visitors own. We say so on the page, and we will keep the caveats next to the charts as they fill in.

The page updates every hour. See the screen stats.