Connect a controller and press any button
Browsers hide gamepads until they receive input (a privacy measure). Pair over USB or Bluetooth, then press any button to wake it.
Plug in a controller, press a button — then measure what your sticks are really doing.
Your files never leave your browser — all processing is local.
Connect a controller and press any button
Browsers hide gamepads until they receive input (a privacy measure). Pair over USB or Bluetooth, then press any button to wake it.
Connect a controller over USB or Bluetooth and press any button to wake it (the Gamepad API only reveals a pad after input — the page guides you through this). Every button lights up on a controller diagram as you press it, triggers show their analog 0–1 values, and multiple controllers can be tested side by side.
The stick analysis is the heart of the tool. A live XY scatter plot with trail shows exactly where each stick sits; a circularity test has you roll the stick around its rim to reveal range problems; a magnified center view makes resting offset visible; and the guided drift test — hands off the controller for 10 seconds — measures average and peak deviation and gives a plain verdict like "left stick drifts 3.2%, above the normal range". Stick drift is the defining fault of modern controllers, and measuring it beats guessing.
Rounding it out: a dual-rumble vibration test with separate strong/weak motor sliders, a polling-rate estimate from timestamp deltas (125 Hz wired pads vs 1000 Hz esports pads show up clearly), and the same rapid re-trigger chatter detection used in our keyboard tester. Everything runs locally in your browser — input data never leaves your device.
Three power-user extras round out the workbench. An input-latency test flashes a prompt and times your press two ways at once: a reaction figure (flash-to-press, which honestly includes your own reaction time) and a pipeline figure that removes human reaction by differencing the Gamepad API's own timestamp against the frame it first becomes readable — the closest a sandboxed page can get to device-plus-polling latency. With two controllers connected, a side-by-side compare mode puts both diagrams and stick scatters in one view, so a known-good pad reads directly against a suspected-drifting one. And an export button downloads the full diagnostic as JSON — id, mapping, button and axis counts, current values and observed min/max ranges, the polling-rate estimate and any drift results — ready to attach to a warranty claim or forum post.
Press any button on it first: browsers hide gamepads until they receive input, as a privacy measure. If it still doesn't appear, check that it's paired/connected at the OS level, and try another USB port or re-pairing Bluetooth. Some very old or exotic pads don't implement the standard mapping and may appear with unusual button numbering.
At rest, a healthy stick should sit within about 1–2% of center (games hide this inside their dead zones, typically 5–10%). Consistent resting offset beyond ~3–5%, or a stick that wanders while untouched, will be visible in gameplay as unwanted movement — that's the classic worn-potentiometer fault that warrants cleaning, recalibration or repair.
It uses the Gamepad API's dual-rumble actuator, which works for most Xbox and PlayStation pads in Chromium browsers. Some browsers and some pads (and most Bluetooth connections on macOS) don't expose rumble — the tool tells you when the actuator isn't available rather than failing silently.
Yes for testing: they pair over Bluetooth and expose the standard mapping in most browsers. Note the browser can only read the controller — it cannot turn it into a system-wide virtual gamepad for other PC games; that requires a driver-level tool like BetterJoy or Steam Input.
How often your controller reports its state, estimated from the timestamps between updates. 125 Hz (8 ms) is the common baseline, high-end wired pads reach 1000 Hz (1 ms). Lower isn't broken — but if a "1000 Hz" pad measures 125 Hz, check its wired connection and mode.
It reports two numbers. The reaction figure (green flash to button press) is dominated by your own reaction time — useful for comparing setups, not for judging a controller. The pipeline figure is more honest about hardware: it measures how stale the pad's state is when a requestAnimationFrame loop can first act on it, by differencing the Gamepad API's timestamp against the frame time. A browser sandbox can't measure true end-to-end display latency, and timestamp granularity plus your polling interval put a floor under the pipeline number, so read it as an upper bound on input freshness rather than a lab-grade figure.
Yes — connect two pads and switch to Compare mode. Each controller gets its own live diagram and stick scatter/drift readout side by side, which is the fastest way to see a worn stick against a healthy one. You choose which two of the connected pads sit in each column.
The selected pad's id, standard-mapping name, button and axis counts, each control's current value plus the minimum and maximum observed while you were testing, the polling-rate estimate, and the drift-test results if you ran them. Everything is generated in your browser and downloaded directly — nothing is sent anywhere. Roll every stick around its rim and squeeze both triggers before exporting so the observed ranges are complete.