Where Does wpt.fyi Get Its Test Data and How Often Is It Updated?

Updated

wpt.fyi (the web-platform-tests dashboard) displays results produced by running the open-source web-platform-tests (WPT) suite in real browsers, then publishing those runs to the dashboard. The data is not hand-entered or editorial: each result set corresponds to a specific test run against a specific browser build, and the dashboard lets you inspect the run's date and browser version so you can judge how current it is. Frequency is not a single fixed schedule — new runs appear as browser vendors and the WPT project publish them, so the practical answer is "check the run metadata for the result set you're looking at."

Where the data comes from

The chain has three parts:

  1. The test suite — web-platform-tests is an open-source, cross-browser test suite maintained by the WPT project (a collaboration involving browser vendors and the wider web community). Tests live in the project's repository and cover web platform features: DOM, CSS, HTML, Fetch, WebDriver, and so on.
  2. The runs — the suite is executed in actual browser engines (Chromium, Firefox, WebKit/Safari, and others). Each execution produces a machine-readable set of pass/fail results tied to a browser name and version.
  3. The dashboard — wpt.fyi ingests those published runs and presents them as comparable result sets, so you can see, for a given feature area, how different browsers performed.

Because the tests are open source, the results are reproducible in principle: anyone can run the suite against a browser build and compare. That is the point of the project — a shared, vendor-neutral yardstick rather than each vendor's own test claims.

How often results are updated

There is no single "updated every N hours" guarantee that applies to the whole dashboard. Instead:

  • Each result set is a snapshot of one run. Its freshness depends on when that run was executed and published, not on when you loaded the page.
  • New runs arrive as they are produced by the participating browser projects and the WPT infrastructure. Different browsers and channels will therefore have different cadences.
  • Stable vs. experimental channels are typically represented separately, so a "latest stable" result and a "development/nightly" result for the same browser can be weeks apart in age.

The practical consequence: never assume the top-line number you see is from today. Always confirm the run's date and browser version before drawing a conclusion, especially when comparing two browsers — an old run against a new one is not a fair comparison.

How to check the date and version of a result set

On wpt.fyi, result views are organized around runs, and each run carries metadata. To verify what you're looking at:

  1. Open the results view for the feature or test directory you care about.
  2. Locate the run selector or the run labels attached to each browser column.
  3. Read the browser version and the run date/time for each column.
  4. If a column looks stale, switch to a more recent run for that browser rather than comparing across mismatched dates.

A quick sanity check before citing any number: are all the columns you're comparing from runs of a similar date, and are you comparing like channels (stable to stable, experimental to experimental)? If not, the comparison is not meaningful.

What this means for reading results

  • Treat each number as "browser X, version Y, run on date Z." That is the full claim the dashboard supports.
  • Prefer recent, matched runs when comparing browsers; the dashboard's value is comparability, and mismatched dates undermine it.
  • Use the underlying test suite for detail. If a result surprises you, the individual test files and their expected results are in the open-source WPT repository, which is the authoritative source behind the dashboard.
  • Expect gaps. A browser or channel with no recent run will simply show older data; absence of a fresh run is not evidence of a regression or an improvement.

In short: wpt.fyi is a window onto published WPT runs, not a live measurement service. Its data is as current as the most recent run for the browser you're inspecting — so check the run's date and version first, and compare only runs of similar vintage.