Aim trainer: method

What it measures

Speed and accuracy of aimed pointing movements with the visitor's own input (mouse, touchpad, pen or finger). Fitts's-law throughput combines movement time and target difficulty in one number (bits per second).

Procedure

A square play area (side = the smallest of the container width, 560 CSS px and 68% of the viewport height) shows a start button in its centre (diameter 20% of the side). Pressing it starts the clock and shows the first target. 30 circular targets then appear one at a time at positions generated on the server from a per-run seed. Diameters are 8%, 11% and 15% of the side, 10 targets each, in shuffled order; each target is at least 25% of the side away from the previous one (the first one from the centre). A pointer-down within the target (radius plus 0.4% of the side) is a hit and shows the next target at once; a pointer-down anywhere else in the area is a miss, and the target stays. No practice trials. Input: left mouse button, touch or pen (pen is recorded as touch). The page records every pointer-down (time since Start, x, y as a share of the side); the server replays these against the seeded targets. Leaving the tab or window during a run interrupts it. A shared challenge link reuses the seed, so a friend gets the same targets.

Parameters

Scoring

Primary metric: mean time per target (ms) = time of the 30th hit since Start / 30; lower is better, and misses add time. Also reported: accuracy = 30 / (30 + misses); median and fastest per-target time. Per target, movement time MT = time from the previous hit (or Start) to the hit, and index of difficulty ID = log2(D/W + 1) bits (Shannon form), with D = distance between the centres of the previous target (or the area centre) and the current target, and W = target diameter. Throughput = mean over targets of ID/MT, in bits per second (higher is better). A least-squares line MT = a + b·ID gives the intercept (ms) and slope (ms per bit). Mean time is also reported per target size. Results are compared only within the same input type.

Valid-run rules

A run is rejected unless all 30 targets were hit, click times never go backwards, and the last hit came within 180 s. A run with a mean time per target under 120 ms is flagged as implausible; it is shown to the player but not counted. Comparisons with other players use first, unflagged runs (one per visitor, account or IP and device) with the same input type, and appear once 100 such runs exist.

Differences from the classic task

Throughput uses the nominal distance and target width of each trial, without the effective-width and effective-amplitude correction of the ISO 9241-9 method described by Soukoreff and MacKenzie (2004), and it is a mean of per-trial ID/MT values rather than a per-condition value. Targets appear one after another at random positions, not in the reciprocal or circular layout of the standard; there are only 30 trials and three widths; distances vary continuously; there is no practice block; a miss does not end the trial.

Limitations

Pointer-event timing in a browser depends on the device, the screen refresh rate and input latency. The play area has different physical sizes on different screens, so distances and widths in millimetres vary. Mouse acceleration and touchpad settings are not controlled. The setting is unsupervised, and times improve over the first runs with practice. Throughput values are not directly comparable with values from the ISO method. Not a clinical or diagnostic measure.

Languages and review status

Language-independent stimuli, instructions in all site languages.