Methodology and limits
Results separate browser observations from user confirmations. Browser checks do not prove SSD SMART, RAM health, CPU/GPU temperature, electrical condition, exact USB polling rate, panel response time, or official certification.
V2 adds ghosting, double-click, wheel, monitor patterns, backlight, refresh estimate, headphones, controller drift, printer, touch, device info, and used-PC / video-call / gaming / new-monitor scenarios. Focused routes reuse the same engines in observation modes. Headphones record hearing only—not audiology or sound-quality certification—and refresh figures are browser frame observations.
Permission denied, unsupported, busy device, skipped, and error stay distinct from a healthy observation. Completion rate is progress, not a probability the device is fine. Local report branding and JSON are this-browser settings only—not signatures or official seals.
Screen patterns advance manually with no automatic high-speed flashing. Speaker and headphone tests use low initial volume and an immediate stop control. Raw pointer paths are not retained.
A My desk mouse summary is the aggregate counts from one run you chose to save. Comparison shows numbers side by side only for the same test type, and flags differences in nickname, connection, observation window, browser, or test time. Differences are only differences between observations — not rates per second, a health score, an improvement percentage, or a defect verdict. Older records reflect the moment they were saved, not the current state. The summary backup JSON is a different format from the guided checkup report JSON.
Practice: typing and precision click scores are a record of how a person performed in this browser, not device check results. Typing counts committed characters only (Unicode NFC code points); characters still being composed by an IME, including one in progress when time runs out, are not counted. Korean uses correct characters per minute (not keystrokes), English uses WPM from correct characters ÷ 5, and accuracy is the share of characters left at the end that match the text. Pasting and dropping are blocked, and rounds where many characters arrive at once or the speed is beyond human typing are not scored. Precision click counts only primary pointer presses inside the area as hits or misses, and labels keyboard or mixed input separately. Both use a monotonic clock for active time only; leaving the window pauses the round until you press Continue. Input lag, polling rate and sensor or panel performance are not measured. Best results are grouped only by challenge, round, language and input method; passage choice, screen and target size, pauses, and a physical versus on-screen keyboard are not held equal, so they are a personal practice reference, not a controlled benchmark or proof of improvement. There is no public ranking.
Signal Arcade games: Word Relay, Route Run, Stereo Echo, Orbit Route, Voice Glide, Color Mosaic and Motion Zones are short games linked to a device check. Their scores count only what the game itself observes (accepted words, checkpoints in order, left/right answers, gates passed, tiles found, zones reached) with a monotonic clock that pauses when you leave the tab. Stereo Echo plays each note in one output channel only and cannot detect whether your device is mono. Orbit Route reads the browser's standard gamepad mapping; other mappings are reported as unsupported, and the keyboard mode is labelled DEMO. Voice Glide compares microphone loudness with a quiet/speaking calibration taken just before the run; it does not recognise speech or judge microphone quality. Color Mosaic depends on your screen, brightness and eyesight and is not a colour-accuracy or vision test. Motion Zones compares small grey preview frames with each other; it does not detect faces, people or gestures and is sensitive to lighting. None of these games measure latency, polling rate, frequency response or panel performance, and they are not device diagnoses.