一格顏色不一樣，點它。愈玩色差愈小，結束時給你一張色感報告。One tile differs in colour — tap it. The gap shrinks each round. At the end, see your colour-resolution report.
每一關的色差不是隨便調 HSL 明度——而是在 OKLCH/OKLab 感知色空間裡，把偏離底色的距離（ΔE）鎖定在固定值。外面那些同類遊戲多半直接把 HSL 明度推幾個百分點，結果同一個「難度」在黃色底色上會變成藍色底色的三倍難——差在哪個色相純屬意外。這裡用 ΔE 鎖定，同一關在六個底色上是同一個距離。底色的色相和擾動軸（色相／彩度／明度）每一關都重新隨機抽——是隨機分散，不是固定輪替，所以連續幾關落在同一個色相是會發生的。Each round's colour gap is not a naive HSL tweak — it's a fixed perceptual distance (ΔE) measured in OKLCH/OKLab space. Most games of this kind just nudge HSL lightness a few percent, which happens to make yellow bases about 3x harder than blue ones — which hue gets punished is an accident of the colour space. Fixing ΔE means the same round is the same distance on all six bases. The base hue and perturbation axis (hue/chroma/lightness) are drawn at random every round — spread by chance, not by a fixed rotation, so a few consecutive rounds can land on the same hue.
六格輪顯示你在各色相分區（紅橙、黃、綠、青藍、藍、紫紅）的辨色門檻——就是你答對率掉到一半的那個 ΔE，越低越敏銳。它是把你歷次遊玩的每一次判斷（色相、實際 ΔE、對或錯）累積起來擬合出來的，不是單獨這一局：單命制下死前每一關都是對的，一局的數字只會反映你死在第幾關，量不到你這個人。所以新玩家會先看到空輪盤，玩幾局才長出數字。某個色相如果你從沒漏過，我們只能說「優於 ΔE X」而不是給一個門檻；斜線格則是那個色相還沒累積夠。螢幕未校準、亮度與環境光未知，這個數字只在你自己的紀錄之間有意義，不能和別人比。The six-sector wheel shows your discrimination threshold per hue sector (red-orange, yellow, green, cyan, blue, violet) — the ΔE at which your success rate falls to half. Lower is sharper. It is fitted from every judgement you have made across runs (hue, actual ΔE, right or wrong), not from a single run: with one life every round before the last is correct by construction, so a single run's numbers only reflect which round you died on, not who you are. That is why a new player sees an empty wheel and it fills in over several runs. If you have never missed in a sector we can only say "better than ΔE X" rather than name a threshold; hatched sectors have not accumulated enough yet. Screen calibration and ambient light are unknown, so the number is meaningful only against your own history, never against someone else's.
顯示的 ΔE 值 = OKLab 歐氏距離 × 100，是一個內部一致、可稽核的難度尺度。它不是 CIEDE2000，我們不宣稱臨床精度。這個遊戲的辨色在未校準的消費者螢幕、未知亮度與環境光下進行，測出來的是「你在這個螢幕這個環境下此刻的辨色力」，而不是你眼睛的生理數值。這不是色覺篩檢，也不能取代眼科醫師的診斷。關於色覺差異，要說清楚：擾動軸與底色每關隨機抽，明度軸（約三分之一的關卡）對紅綠色覺差異幾乎無影響；但色相軸在黃色與藍色底色上，數學上就是沿著紅綠方向移動，對紅綠色弱的人這些關會明顯更難。所以這個遊戲對紅綠色弱並不是完全公平的，報告裡的「依差異種類」那一欄才是對你有意義的讀數——輪盤那六個色相是把三種差異混在一起算的。Displayed ΔE = OKLab Euclidean distance × 100 — an internally consistent, auditable difficulty scale. It is not CIEDE2000 and we make no clinical precision claim. Discrimination here happens on an uncalibrated consumer screen at unknown brightness and ambient light; what you measure is "your ability on this screen in this room right now," not a physiological eye value. This is not a colour-vision screening and cannot substitute for a clinician's examination. On colour-vision differences, plainly: axis and base hue are drawn at random each round, and the lightness axis (about a third of rounds) is essentially unaffected by red-green differences. But a hue-axis round on the yellow or blue base moves along the red-green direction almost exactly, so those rounds are genuinely harder if you have a red-green deficiency. This game is not fully fair in that respect. The "by kind of difference" rows in the report are the readout that means something for you — the six hue sectors on the wheel pool all three kinds together.
Quick guide
Compact below-tool notes that help first-run users and repeated visitors move faster without changing the main interface.
Chinese search: 找不同顏色、色塊遊戲、色感測試、眼力測試、分辨顏色遊戲、一格之差、色相辨色、色覺遊戲
How to use
Examples
FAQ
No. This is a rough personal measurement on an uncalibrated screen at unknown brightness and ambient light — it's a game. Real colour-vision screening requires a clinician, not a web page.
Difficulty is a perceptual colour difference (ΔE) in OKLCH/OKLab space, not a raw HSL lightness tweak. Most games of this kind nudge lightness a few percent, which happens to make yellow bases roughly 3x harder than blue ones — which hue gets punished is an accident of the colour space. Fixing ΔE means round N is the same distance on all six bases.
Every judgement you have made across runs (hue sector, the ΔE actually achieved, right or wrong) is accumulated, and each sector is fitted for the ΔE at which your success rate falls to half. Why across runs? With one life, every round before the last is correct by construction, so a single run's numbers only reflect which round you died on — they cannot measure you. A sector you have never missed is reported as "better than ΔE X" rather than given a threshold; sectors without enough data stay hatched instead of pretending.
Displayed ΔE = OKLab Euclidean distance × 100. It's an internally consistent, auditable difficulty scale, not CIEDE2000. We make no clinical precision claim.
Partly, and it is better to say so plainly. The axis and base hue are drawn at random each round, and the lightness axis — about a third of rounds — is essentially unaffected by red-green differences. But a hue-axis round on the yellow or blue base moves almost exactly along the red-green direction, so those rounds are genuinely harder with a red-green deficiency. The game is not fully fair in that respect. The "by kind of difference" rows in the report (hue / chroma / lightness) are the readout that means something for you; the six hue sectors on the wheel pool all three kinds together.