Huefork

What Huefork trains

Huefork does two jobs in one round. First it asks whether you can detect a shade difference at all, spotting the odd swatch among three. Then it asks whether you can quantify it: was that swatch lighter or darker, and by a lot or a little. Detection alone is not much use in practice: noticing that two greys disagree only helps if you can also say which way to move and roughly how far. Across five rounds the difference shrinks step by step, closing in on the smallest lightness gap you can still call correctly.

How to play

Each round shows three swatches. Two are identical; one differs in lightness only, at the same hue and the same saturation. Step 1: tap the one that is different. The true odd swatch is then marked whether you found it or not. Step 2: rate how different it looked on a five-point signed scale, from much darker through barely any difference to much lighter, and press Confirm estimate; if your rating is not exact, the correct notch is highlighted before you move on. There are five rounds and no clock. The differences shrink each round, running roughly 20, 12, 8, 5 and 3 lightness units, so the last round sits near the limit of what a screen can show you. The scale is a proper radio group: arrow keys move between notches, Enter or Space selects.

How scoring works

Each round is worth 100 points, split three ways: 50 for finding the odd swatch, 25 for the direction (lighter or darker), and 25 for the magnitude. Magnitude only counts if the direction is right: an exact match scores the full 25, and one notch out scores 12. The headline number is not that total, though. The result card leads with your threshold: the smallest difference of the day at which you still got the direction right, shown as a ΔL figure, or as "greater than" the starting difference if even the widest round went the wrong way. A threshold of 8 or better earns the day's streak.

What your score means

The threshold number is the one to watch, and smaller is better. It can only ever be one of the five differences the day actually used (20, 12, 8, 5 or 3), or "greater than 20" if no round went the right way:

Your threshold depends heavily on screen brightness and the room, so treat it as your number on your setup, and follow its trend rather than its absolute value.

Common mistakes

The most costly habit is guessing the direction after a missed detection. If you did not find the odd swatch, the game still shows you which one it was, so look at it before rating, because direction and magnitude are worth half the round and they are still winnable. Second, players use only the extreme notches; on the later rounds the honest answer is the slight one, and the exact notch is worth double the near miss. Third, staring: prolonged fixation on a flat field lets the difference fade, so a quick comparison, a look away and a second glance works better. Finally, treating the middle notch as a safe answer costs both direction and magnitude, since it commits to no direction at all.

Practice progression

(5 min, beginner) On each round, say which way before you rate how far, because locking the direction first protects the points that matter most. (10 min, intermediate) In the practice tab, set the starting difference low so every round sits near your limit, and add rounds so the shrinking sequence goes further down. (weekly) Pair it with Hueshort, which trains the same discrimination under time pressure instead of at the limit, and read Delta-E & the Just-Noticeable Color Limit.

The science behind it

Picking the odd one from three is a classic forced-choice design: because you must choose, guessing alone gets you one in three, and performance above that reflects real sensitivity rather than confidence or a willingness to say "yes". Shrinking the difference round by round and finding where you stop being right is the idea behind a threshold measurement, the point at which a difference becomes just noticeable. The second half of each round is magnitude estimation: judging not whether two things differ but by how much, a separate and noisier ability. Thresholds are also not fixed; they depend on the surrounding field, on display brightness, and on the base lightness itself. See Delta-E and the just-noticeable limit and Simultaneous Contrast for why context moves the answer.

Limitations

Huefork is a game and an informal self-check, not a clinical or diagnostic tool. Your reported threshold is a browser measurement on an uncalibrated screen in unknown light, estimated from five trials, far too few to separate a real change in sensitivity from luck. It tests lightness differences only, at one hue per round, and says nothing about hue discrimination or color vision as a whole. Compare it against your own past results on the same device, and against nothing else.

Where it shows up

Detect-then-quantify is the everyday shape of color work. A print operator notices a sheet has drifted, then has to say whether to add or pull density and by how much, because noticing alone does not move the press. A retoucher sees that a patched area does not sit right and must decide the direction and size of the correction rather than nudging blindly until it stops looking wrong. A painter matching an existing wall sees the test patch is off and needs to know whether to add white or the darker pigment, and roughly how much, because each extra pass costs a drying time. A dentist choosing a shade for a crown, a technician reading a colorimetric test strip, a tailor matching a replacement panel to a faded garment: all of them face the same two-stage question. Quality control works this way too: the useful report is not "this one differs" but "this one is two shades light". Detection tells you something is wrong; magnitude and direction tell you what to do about it, and only the second part actually fixes anything. Huefork scores both, in that order, which is why a missed detection still leaves half the round on the table.

Related reading & games

• Delta-E and the just-noticeable limit: what a threshold number means and how it is measured.
• Simultaneous contrast: why the same swatch looks different against a different background.
• Value and saturation: why lightness carries most of what you see.
• Hueshort (the same call against a clock) and Hueseam (find a faint edge in a wide bar): related threshold games.
• Learn about color: the full guide library.