A reaction-time score includes latency from the display, the operating system, the input device, and the browser. Some of that latency is fixed by the hardware; some of it is a setting you can change. The practical goal is not to eliminate every millisecond — that is not possible in a browser — but to know which settings matter, keep them consistent, and avoid fooling yourself when you change something.
What you can and cannot measure at home
At home you can measure relative changes: the same test, same device, same conditions, before and after one setting change. You cannot easily measure the absolute photon-to-click latency of your system without specialized equipment. A browser test reports the interval between its own start signal and the input event it receives; it does not know when the pixel actually lit up or when your finger first moved.
That is why this guide focuses on controlled comparisons rather than absolute numbers. A change that consistently improves your median over several sessions is worth keeping. A single fast round after a settings change is not evidence of anything.
Start with the display
Check that the display is running at its intended refresh rate. Operating systems sometimes default to 60 Hz even on a high-refresh monitor, especially after a cable or dock change. The setting is in the display or graphics control panel; a browser cannot reliably tell you the panel's true refresh rate. If the desktop feels less smooth than expected, check this first.
Then look at the display's processing modes. Many TVs and some monitors have a "game mode" that reduces image processing; motion interpolation, noise reduction, and scaling can add latency. VRR (variable refresh rate) can reduce stutter and tearing, but its effect on input latency depends on the implementation and the frame rate. If you are comparing settings, change one at a time and test under the same content.
Check the input path
For a mouse, the polling rate determines how often the device reports its state: 125 Hz reports every 8 ms, 500 Hz every 2 ms, and 1,000 Hz every 1 ms. Higher polling can reduce the input delay slightly, but it also increases CPU load and may not be stable on every system. Wireless mice can be excellent, but interference, battery state, and the receiver's placement can add variability; a wired connection removes one variable if you are troubleshooting.
Mouse debounce settings can also matter. A very short debounce may register faster presses but can produce unintended double clicks; a long debounce can filter out very fast repeated presses. If you change this setting, test it in normal use as well as in the reaction test, and keep it consistent for comparisons.
Browser and operating-system settings
Hardware acceleration, power mode, background apps, and browser extensions can all affect rendering and event scheduling. A laptop in a power-saving mode may reduce CPU or GPU clocks and increase latency. Closing heavy background work and using the same browser profile for each session reduces noise. Avoid measuring while the machine is installing updates, indexing files, or running a video call.
Do not assume that every "gaming" setting helps. Some options change visual quality rather than latency; others may be disabled by the browser or the operating system. The only reliable test is a controlled A/B comparison on your own machine.
A simple A/B protocol
To test whether a setting change helps, use the reaction test as your measurement instrument:
- Baseline. Run three rounds under your current settings, at a similar time of day, and record the median of the five-attempt averages.
- Change one setting. For example, enable game mode or switch the mouse to a different polling rate. Change nothing else.
- Re-test. Run three rounds under the new setting, under the same conditions.
- Compare medians. A difference smaller than your normal round-to-round variation is not evidence; repeat on another day before keeping the change.
This is a small experiment, not a laboratory study. Its value comes from controlling the obvious variables: same test, same device, same time, one change at a time. If you change three settings at once, you will not know which one mattered.
The slow-motion camera method
If you want a rough external check, you can film the screen and your hand with a phone in slow-motion mode. The idea is to capture the moment the signal appears and the moment your input registers, then compare timestamps. This is approximate: the phone's frame rate, rolling shutter, exposure, and the difficulty of aligning the two events limit the accuracy. It can reveal a large, obvious delay — for example, a display processing mode adding a visible lag — but it is not a substitute for proper measurement equipment.
Use it as a sanity check, not as a calibration standard. If the slow-motion clip shows a large gap that matches a setting you can change, that is useful information; if it shows a few milliseconds of difference, do not over-interpret it.
A quick checklist before you calibrate
Before changing anything, confirm the basics: the display is set to its native resolution and intended refresh rate, the browser window is not being scaled, the machine is plugged in and not in a power-saving mode, and no heavy download or update is running. A surprising number of "latency" problems disappear when the display is accidentally set to 60 Hz or the browser is running a huge extension load.
Write down the baseline conditions before you touch a setting. If you do not know what "before" was, you cannot tell whether the change helped. A photo of the settings screen is enough; the goal is a record, not a report.
Why your score can change with no setting change
Reaction-time scores vary for reasons unrelated to latency. Sleep, caffeine, time of day, attention, and familiarity with the task all move the number. A noisy background process can cause a few slow attempts. A single change in room temperature or posture can shift your hand's comfort and consistency. This is why a before/after comparison needs several rounds and, ideally, a repeat on another day.
Device latency is also not perfectly constant. The operating system schedules work, the browser handles events in a shared process, and the display's processing can vary with content. A setting that helps on average may still produce round-to-round noise. Look for a consistent shift in the median, not a single impressive attempt.
When to stop calibrating
Calibration has diminishing returns. Once you have a stable setup, a clean browser profile, a sensible display mode, and a consistent input path, further tweaking is more likely to add noise than to improve your score. The remaining milliseconds are increasingly hard to attribute, and the time is better spent on sleep, practice, or simply using the test consistently.
Stop when the change you are testing is smaller than your normal variation, or when you cannot reproduce the result on a second session. Chasing a one-millisecond difference in a browser measurement is not a good use of time. A practical rule is to stop when your setup is stable and your scores are reproducible; from there, the bigger gains come from sleep, practice, and consistency rather than another toggle.
What usually matters, and what usually does not
Usually matters: display refresh rate and processing mode, a stable input connection, enough CPU/GPU headroom to render consistently, and a browser/OS that is not overloaded. Usually does not matter much for a simple reaction test: mouse DPI (it changes cursor distance, not the click latency), RGB lighting, premium cables within spec, and small differences in mouse shape unless they change your grip and comfort.
Comfort is not a soft factor. A setup that makes your hand tense or your wrist ache will produce worse, more variable results over a session. Ergonomic changes can matter more than a millisecond of theoretical latency.
What to record
Keep a short setup note: display model and refresh rate, connection type, mouse polling rate, browser, OS, and any relevant mode (game mode, power mode, hardware acceleration). Without that note, a score from last month is not comparable to today's. When you change something, record the date and the change, then re-baseline.
For the underlying measurement chain, read what a reaction time score really includes. For device-class comparisons, see mouse vs touchscreen reaction time. For the test itself, use the reaction time test.
Limits
A browser cannot measure your panel's true latency, your mouse's sensor-to-click delay, or the exact moment a pixel changes. It measures the interval between its own events. That makes it useful for relative comparisons and personal tracking, and unsuitable for claims like "my setup has exactly 12 ms of lag". Treat the numbers as a practical instrument, not a laboratory certificate.
What home calibration cannot tell you
Home tests can reveal large differences between devices and settings, but they cannot separate panel processing from compositor delay, input firmware, or browser scheduling with laboratory precision. Use the comparison to spot big outliers, not to claim a millisecond-accurate calibration. If a device looks unusually slow, re-test it against a known-good setup before replacing hardware.
For the reaction-time measurement side, see the reaction time guides.
Sources and tools
- MDN Web Docs. Performance.now() and requestAnimationFrame().
- Reaction time test accuracy — the full display/input/browser chain.
- Mouse vs touchscreen reaction time — why device classes differ.
- Methodology — how ReflexTimer builds and scores its tests.