How We Test

Every verdict on this site comes from a published formula and a fixed set of limits. This page shows how each number is calculated, what a browser can't measure, and every change we've made to the limits.

Updated Nov 18, 2026Limits v1.0About 7 minutes to read
On this page
  1. The short version
  2. Where the numbers come from
  3. Stick drift
  4. Circularity
  5. Deadzone
  6. Triggers
  7. Buttons
  8. Vibration
  9. Report rate
  10. What a browser can't measure
  11. All limits
  12. Change log

The short version

Your browser's own readings

Every number comes from the gamepad data your browser already gives web pages. No downloads or special drivers.

Fixed, published limits

Every controller is judged against the same limits. They're all listed below, along with every change.

Nothing leaves your device

All the maths runs on the page. A share link carries the result inside the link itself.

Where the numbers come from

Your controller sends its state to your computer or phone. The system and its driver turn that into standard values, and the browser hands them to web pages through the Gamepad API. From that, we read three things:

  • Axes. Each stick has an X and a Y axis, from −1 to 1. Zero on both means dead center.
  • Buttons. Each button reads from 0 (not pressed) to 1 (fully pressed). Triggers are buttons that can sit anywhere in between.
  • Timestamp. When the browser last received new data from the controller.
ControllerSensors in the sticks, triggers and buttons
System and driverTurn the signal into standard values. Some apply a small deadzone.
BrowserChecks the controller on its own schedule and shares the values
This pageCalculates every number you see. Nothing is sent anywhere.
axes[0–3]buttons[0–17]timestamp
Figure 1. We read values after your system has processed them. That's close to what games on the same device see, but it isn't the raw sensor signal.

Stick drift

Drift is a stick reporting movement while nobody is touching it. Here's how the drift test and the health check measure it:

  1. You let go of both sticks.
  2. For 3 seconds, we record every reading of each stick's X and Y.
  3. For each reading, we work out its distance from the center.
  4. The resting distance is the median of those distances.
// one reading d = √(x² + y²) // over 3 seconds resting distance = median(d) spread = max(d) − min(d) deadzone to hide = ceil((max(d) + 0.02) × 100) / 100

We use the median so that one twitch doesn't change the verdict. We don't subtract a resting offset, and we don't apply a deadzone of our own.

0.10 0.25
Each dot is one reading taken while the stick was left alone.
The ring marks the median position. Its distance from the center, here 0.138, is the resting distance.
Dashed rings are the 0.10 and 0.25 limits. This stick lands between them, so it gets a warning.
Figure 2. A right stick with mild drift. The plot is zoomed in: its edge is 0.30, not 1.
Pass under 0.10Warn 0.10 to 0.25Fail 0.25 and upSet

These starting limits follow the open-source stick-drift-detect project (MIT licence). After testing more controllers we may set different limits per model, and every change will be listed in the change log.

Circularity

A stick pushed to its edge should reach the same distance in every direction.

  1. You rotate the stick slowly around its edge, 3 full turns.
  2. We split the circle into 72 slices of 5° each and keep the farthest reading in each slice. That's the stick's reach in that direction.
  3. The average error is the mean gap between each slice's reach and a perfect 1.0.
reach(slice) = max(d) in that 5° slice average error = mean(|1 − reach|) over all 72 slices
The dashed circle is a perfect reach of 1.0 in every direction.
One 5° slice. The dot is the farthest reading in it.
The blue shape joins every slice's reach. This one is slightly square, which is common.
Figure 3. Many controllers reach a little past 1.0 at the diagonals and a little short of it straight up or sideways. The shape here is exaggerated so it's easy to see.
Pass average error under 8%Warn 8% and upProvisional

We'll confirm this limit after testing more controllers. Until then, circularity never produces a fail.

Deadzone

The deadzone test suggests in-game settings for one stick at a time. It gives suggestions, not a verdict.

  1. You let go for 2 seconds. We record the largest distance from the center: the resting max.
  2. You push the stick slowly to the edge. The first reading clearly above the resting max is the first movement. If the stick read exactly 0 at rest and its first movement is above 0.04, the controller has a built-in deadzone.
  3. You turn the stick once around the edge. As with circularity, we keep the farthest reading in each of 72 slices.
inner deadzone = max(0.03, ceil((resting max + 0.02) × 100) / 100) // none needed when the controller has a built-in deadzone outer deadzone = floor(lowest reach × 100) / 100 // none needed when the lowest reach is 0.99 or more
Suggestions only no pass or failProvisional

Triggers

Triggers are analog. While you press each one all the way and let go, we record its lowest and highest value.

Pass rests at 0.02 or below and reaches 0.98 or aboveWarn anything elseProvisional

A trigger that stops short of the top may never register a full press in some games.

Buttons

During the health check, you press each button once. A button passes if it registers. We flag it as stuck if it reads as pressed before you touch it, and as bouncing if a single press registers twice within 40 ms.

Pass registers cleanlyWarn stuck or bouncingFail no responseProvisional

Vibration

A web page can ask the browser to vibrate the controller, but it can't tell whether the motors actually spun. So we play each motor and ask whether you felt it. The result is felt, not felt or not supported in this browser.

“Not supported” is a browser limit, so it never counts against your controller.

Report rate

We count how many times per second the timestamp changes, which tells us how often the browser receives new data from the controller.

The number is capped by how often the browser itself checks the controller, and it only changes when the input changes, so keep a stick moving while it measures. It is not the controller's own polling rate, and it is not input lag.

We also rate how steady the readings are, from the jitter: how much the gap between readings varies, compared with the average gap.

Steady jitter under 15% of the average gapSome jitter under 35%Uneven 35% and upProvisional

What a browser can't measure

We'd rather tell you what's out of reach than show a number that looks precise and isn't.

  • The controller's real polling rateBrowsers check controllers on their own schedule, which is often slower than the controller can report.
  • Input lagTiming a press all the way to the screen needs a camera or dedicated hardware, not a web page.
  • Readings before the driverSome systems apply a small deadzone of their own, which can hide very small drift.
  • Adaptive triggers, light bar, motion sensors and batteryThese need a lower-level browser feature that this site doesn't use yet.

All limits

The full set of limits behind every verdict. Provisional limits may change after more hands-on testing.

CheckPassWarnFailStatus
Stick drift (resting distance)< 0.100.10 – 0.25≥ 0.25Set
Circularity (average error)< 8%≥ 8%—Provisional
Deadzone (suggested settings)Suggestions only, no verdictProvisional
Trigger range≤ 0.02 → ≥ 0.98anything else—Provisional
Buttonsregisters cleanlystuck or bouncingno responseProvisional
Vibrationfeltnot felt—Set
Report rate (steadiness)Steady < 15%Some jitter < 35%Uneven ≥ 35%Provisional

Change log

  1. Limits v1.0

    First published limits. Stick drift limits follow stick-drift-detect. Circularity, deadzone, trigger, button and report rate limits are provisional.

  2. Next

    Per-model limits, once we've tested enough DualSense and Xbox controllers.

Does a result look wrong for your controller? Tell us which controller, browser and system you used, and we'll look into it.

Contact us