測量方法與限制Methods & Limits

動態殘影與更新率（UFO 測試）Motion blur and refresh rate (UFO test)

怎麼量：以瀏覽器每次重繪的間隔（requestAnimationFrame）顯示網頁實際收到的更新率；殘影由你用眼睛比較。
量不到：面板的反應時間（GtG）、過衝、背光閃爍。
注意：省電模式、背景分頁與可變更新率（VRR）都會改變數字，請在前景分頁、插電狀態下測試。How: uses the interval between browser repaints (requestAnimationFrame) to show the refresh rate the page actually receives; blur is judged by your eyes.
Cannot measure: panel response time (GtG), overshoot, backlight flicker.
Note: power saving, background tabs and variable refresh rate (VRR) change the number; test in a foreground tab while plugged in.

螢幕數據健檢Monitor Health Check

怎麼量：用畫面更新間隔的中位數得到更新率，用螢幕尺寸與縮放比例換算輸出解析度，再與我們內建的型號規格庫比對。
量不到：面板真正的規格是否與廠商標示一致，也無法判斷線材頻寬。
注意：規格庫由人工整理，可能有誤或缺漏，歡迎用「規格有誤？」回報。How: takes the median frame interval for refresh rate, converts screen size and zoom to output resolution, and compares them with our built-in model spec list.
Cannot measure: whether the panel truly matches the vendor's specification, or cable bandwidth.
Note: the spec list is compiled by hand and may be wrong or incomplete; please report mistakes with "Specs wrong?".

壞點與 OLED 烙印檢測Dead pixel and OLED burn-in checks

怎麼量：顯示全螢幕純色供你用眼睛檢查；烙印檢測用引導畫面與自我評估，並可選擇用手機拍照做亮度均勻度分析（照片只在你的瀏覽器處理，不會上傳）。
量不到：是否符合保固條款（依各品牌而定）。
注意：照片會受環境光、相機自動曝光與角度影響，只能當作「有沒有可疑區域」的參考。How: shows full-screen solid colors for visual inspection; the burn-in check uses guided screens and self-assessment, with optional phone-photo uniformity analysis (processed only in your browser, never uploaded).
Cannot measure: whether a defect meets a warranty policy (it depends on the brand).
Note: photos are affected by room light, camera auto-exposure and angle, so treat them as a hint of suspicious areas.

色域估算與對比度Gamut estimate and contrast

怎麼量：讀取瀏覽器回報的色域與 HDR 能力，再用色階圖讓你的眼睛找出顏色不再變化的位置，據此估算覆蓋率。
量不到：螢幕實際發出的光，因此無法取代色度計。
注意：覆蓋率是估算值，受瀏覽器色彩管理、作業系統設定與你的視覺影響。How: reads the gamut and HDR capability the browser reports, then uses a color staircase so your eyes find where colors stop changing, and estimates coverage from that.
Cannot measure: the light the screen actually emits, so it cannot replace a colorimeter.
Note: coverage is an estimate affected by browser color management, OS settings and your own vision.

VRR 與幀率穩定度VRR and frame pacing

怎麼量：記錄一段時間內的畫面間隔分布與掉幀次數。
量不到：顯示卡驅動中 G-Sync／FreeSync 是否真的在硬體層生效。
注意：網頁本身的負載與瀏覽器排程也會造成卡頓，不一定是螢幕的問題。How: records the frame-interval distribution and dropped frames over time.
Cannot measure: whether G-Sync / FreeSync is actually active in the GPU driver and hardware.
Note: page load and browser scheduling can also cause stutter, so it isn't always the monitor.

耳機延遲（手動對位）Headset latency (manual alignment)

怎麼量：你調整滑桿，讓畫面上的球落地與聽到的聲音對齊，滑桿位置就是延遲估計。
量不到：客觀的精確數值，結果取決於你的判斷。
注意：適合在同一副耳機上比較不同連線方式。How: you move a slider until the ball landing on screen lines up with the sound you hear; the slider position is the latency estimate.
Cannot measure: an objective precise value; the result depends on your judgement.
Note: best for comparing connection types on the same headset.

迴路延遲自動測量（RTL）Automatic loopback latency (RTL)

怎麼量：播放一個 1 kHz 短嗶聲，同時用麥克風收音，計算「發出」到「收到」的時間。
量不到：耳機單向的輸出延遲。這個數字包含瀏覽器與作業系統的音訊處理、輸出裝置（有線、藍牙或 2.4G）、聲音在空氣中的傳播與麥克風輸入。
注意：只適合拿來比較「同一套設備在不同設定下」的差異；不同麥克風、不同電腦之間的數字不能直接比較。排行榜是玩家自行上傳的結果，僅供相對參考。How: plays a short 1 kHz beep while recording with the microphone and calculates the time from "emitted" to "received".
Cannot measure: the headset's one-way output latency. The number includes browser and OS audio processing, the output device (wired, Bluetooth or 2.4G), sound travel through air and the microphone input.
Note: only useful for comparing the same setup under different settings; numbers from different microphones or computers are not directly comparable. The leaderboard is player-submitted and for relative reference only.

頻譜分析與 AI 麥克風診斷Spectrum analyzer and AI mic check

怎麼量：用 Web Audio 取得即時頻譜，並在瀏覽器內執行聲音辨識模型，再依音量、削波、頻寬與電流聲等指標給出參考評分；音訊不會上傳。
量不到：麥克風本身的絕對品質。
注意：評分受距離、房間噪音與系統增益影響，分數是參考，不是認證。How: uses Web Audio for a live spectrum and runs a sound-recognition model in the browser, then gives a reference score from level, clipping, bandwidth and hum; audio is never uploaded.
Cannot measure: the absolute quality of the microphone itself.
Note: the score depends on distance, room noise and system gain; it's a reference, not a certification.

聲道、相位、聽聲辨位、頻率掃描與煲機工具Channel, phase, localization, sweep and burn-in tools

怎麼運作：它們是訊號產生器，由你用耳朵判斷。
注意：煲機是否真的改變耳機音質仍有爭議，我們把它當成可選的訊號播放工具，不保證效果；請使用中低音量，感到不適就立即停止。How: these are signal generators that you judge by ear.
Note: whether burn-in really changes headphone sound is disputed, so we treat it as an optional signal player with no guaranteed effect; use a moderate volume and stop if it is uncomfortable.

滑鼠回報率Mouse polling rate

怎麼量：計算短時間內瀏覽器收到的滑鼠移動事件數量（支援時使用原始指標事件）。
量不到：USB 裝置實際送出的原始封包速率。
注意：事件可能被合併，瀏覽器與作業系統的排程也會讓數字偏低，4000／8000 Hz 滑鼠尤其明顯；請快速移動滑鼠，並以同一瀏覽器重複比較。How: counts the mouse-move events the browser receives in a short window (using raw pointer events when supported).
Cannot measure: the raw packet rate the USB device actually sends.
Note: events can be coalesced, and browser and OS scheduling lower the number, especially for 4000 / 8000 Hz mice; move the mouse quickly and compare using the same browser.

鍵盤延遲與 NKROKeyboard latency and NKRO

怎麼量：記錄按鍵事件的時間與同時被偵測到的按鍵數。
量不到：鍵盤從開關到 USB 的完整硬體延遲。「按鍵到畫面」是包含顯示延遲的相對比較值。
注意：作業系統與瀏覽器會攔截部分快捷鍵，連線模式與韌體也會影響能同時送出的按鍵數。How: records key-event timestamps and how many keys are detected at once.
Cannot measure: the full hardware latency from switch to USB. "Input-to-frame" is a relative figure that includes display latency.
Note: the OS and browser intercept some shortcuts, and connection mode and firmware affect how many keys can be sent at once.

搖桿死區與飄移Controller deadzone and drift

怎麼量：透過 Gamepad API 讀取搖桿軸的數值，顯示靜止時的偏移與死區範圍。
量不到：硬體磨損程度。
注意：不同瀏覽器、連線方式與驅動回報的數值可能不同；請在放開搖桿、靜置狀態下測試。How: reads the stick axis values through the Gamepad API and shows resting offset and deadzone.
Cannot measure: the degree of hardware wear.
Note: values can differ between browsers, connection types and drivers; test with the stick released and at rest.

反應時間與跟槍Reaction time and tracking

怎麼量：用高精度計時器記錄畫面變化到你按下的時間，多回合取平均。「估計肉體反應」是從總成績扣掉依你選擇的設備推算的硬體延遲，是假設性的模型，不是儀器量到的神經反應時間。
注意：疲勞、練習與預判都會影響成績，建議多測幾次、在固定條件下比較。How: uses a high-resolution timer for the time from the visual change to your press, averaged over several rounds. "Estimated human reaction" subtracts a hardware latency inferred from the gear you selected; it is a hypothetical model, not an instrument reading of nerve reaction time.
Note: fatigue, practice and anticipation affect results; repeat under the same conditions.

3D 練槍、MOBA 操作測試、打字速度3D Aim Trainer, MOBA APM Test, typing

怎麼計分：依遊戲內的命中、閃避、擊殺或正確字數計算。
排行榜：伺服器會依提交的計數檢查成績是否合理，並可隱藏明顯不合理的成績。
硬體報告：只列出實際量到的項目，例如畫面更新率、掉幀與可量得的輸入事件頻率。Scoring: based on in-game hits, dodges, kills or correct characters.
Leaderboards: the server checks submitted counters for plausibility and may hide clearly implausible scores.
Hardware report: lists only items that were actually measured, such as refresh rate, dropped frames and measurable input-event rate.