What compatible browsers expose
ScrollTest depends on the browser's standard wheel-event interface. A received event can include horizontal and vertical deltas, a delta mode, modifier-key state, and a high-resolution event timestamp. ScrollTest applies its own monotonic timing at capture so elapsed intervals use one deterministic clock.
Browsers may coalesce, schedule, or transform events. Operating systems and device drivers may apply acceleration, sensitivity, momentum, or application-specific behavior before the page receives anything. Those layers are why event rate is labelled browser events per second and why reported delta retains its pixel, line, or page unit.
Desktop browser differences to expect
| Variable | Possible effect | How to handle it |
|---|---|---|
| Event scheduling | Intervals can cluster or stretch under load | Close heavy work and repeat; do not infer hardware from one burst |
| Delta mode | Movement may be reported as pixels, lines, or pages | Compare only like units; never relabel them as pixels |
| Zoom and page scale | CSS-pixel interpretation and page behavior can change | Keep zoom fixed for a repeatability series |
| Extensions | Scrolling can be intercepted, smoothed, or remapped | Retest in a clean browser profile when results differ |
| Main-thread load | Delivery may be delayed even when input is steady | Review callback delay; more than 250 ms invalidates the run |
Background tabs and visibility changes
Browsers deliberately throttle timers and work in hidden tabs. Switching applications, minimizing the window, or moving to another tab can therefore corrupt a timed sample. ScrollTest listens for document visibility changes and flags or invalidates an active run. The actual elapsed duration remains part of the result; a nominal 10-second choice does not justify hiding a delayed finish. A timer callback delay of 250 ms or less remains valid and is saved with the result. A delay over 250 ms is material, invalidates the run, and excludes it from history and comparisons.
Keep the test zone focused and visible until completion. If an interruption is reported, start a fresh run rather than comparing the interrupted trace with local history.
Touchscreens are not touchpads
A two-finger touchpad gesture is commonly translated by the OS and browser into wheel events. A finger swipe directly on a touchscreen is usually exposed through touch or pointer behavior and may scroll the page without producing the same wheel-event stream. ScrollTest is a wheel- event diagnostic, so a touchscreen swipe is not a substitute for the declared touchpad category.
Mobile browsers may also reserve gestures for navigation or viewport control. The interface can remain readable on a phone or tablet, but the core wheel test needs a connected device or input path that actually generates wheel events. Read the mouse and touchpad comparison for category guidance.
Private browsing and local storage
Repeatability history is stored in the browser. Private modes, storage blocking, clearing site data, or browser eviction can remove it or prevent it from being saved. A test can still run when storage is unavailable, but history and personal comparison features may not persist.
Browser sync is not used to move ScrollTest history between devices. An explicit export is the portable copy. See the privacy policy for retention and deletion controls.
A practical cross-browser check
- Record the device category, input-device connection, wheel mode, OS settings, and browser.
- Complete the same controlled direction protocol at least three times, including normal and faster paces.
- Open a second current browser without changing device or system settings.
- Repeat the same protocol and compare units before comparing values.
- Interpret a difference as environment evidence, not as a browser accuracy ranking.
If one environment repeatedly shows a symptom, use the troubleshooting sequence to change one variable at a time.