Velocity thresholds must be recalibrated across size and resolution

What it is

Copying 800 px/s from a phone to a tablet or a 4K car display makes swipes all fail or all too eager. The threshold must be retuned in physical speed (mm/s) or viewport fractions (screen-widths per second). Pixel rate is not device-independent. The same arm motion is a different pixel derivative on higher PPI or a larger sheet of glass.

Why it happens

Pixel speed = physical speed × PPI. Raise PPI and the same flick looks “faster” in pixels, so a fixed px/s cut promotes ordinary pans to swipes. On a larger screen people use a longer arc for “one step”; keeping the phone’s mm/s then makes flicks dull. Viewport-relative units bind “how violent is a page” to the current container width, which is stabler for pagers. Cars and TVs are viewed from farther away, so the same angular speed of the arm is more mm/s on the glass—yet another conversion. What is calibrated is the classifier’s unit, not the inertia damping coefficient.

Studying it

The same people do “stop halfway / fling next” on a phone, a tablet, and a large car screen. Fit thresholds in px/s, mm/s, and screen-widths/s, and compare match rate after cross-device transfer. Transfer error is the measure that matters: how far a phone-tuned px/s drifts on a tablet. Do not tune on one emulator resolution.

Where it stops holding

Mouse wheels and trackpads are other transducers and do not take touchscreen mm/s. Window resize and split-screen change viewport width; a width-relative threshold must follow the live container, not the factory device size. Game world speed sits behind a camera. Recalibration removes unit error; it does not remove chatter at the cut.

Applying it

• Store thresholds as mm/s or container-widths/s, and multiply by current PPI and viewport at runtime.
• Measure a ROC on phone portrait, tablet landscape, and a desktop window; forbid one pixel constant.
• Drop the phone-tuned number onto a tablet unchanged: if flicks become trivial or impossible, the unit was not converted.

Related

• Same group: C3.29.1 The swipe-versus-pan velocity cut decides a jump or a follow · C3.29.2 Velocity must be judged together with distance · C3.29.3 Near the cut, the same stroke flickers between two readings
• Adjacent: C2.17 Touch sampling rate and report rate · C3.06 Fling and inertia
• Search: physical velocity · PPI · viewport-relative threshold