Spot the different colour game

Colours are drawn only when Start is tapped, never before, and every tile is equally likely to be the odd one. A tap on the right tile clears the level and deals the next grid in new colours. A tap on any other tile is a mistake, and the way of playing decides what it costs.

• Against the clock: the timer starts at 60 seconds, and a wrong tap takes 3 seconds off it while the same grid stays on screen.
• Three lives, no clock: a wrong tap costs a life, the odd tile is ringed for a moment, and the level is dealt again in new colours. Losing the last life ends the game.
• Today's colours: the clock rules again, but each level's colours and odd tile are identical for everyone who plays on that calendar date with the same option.

No organisation sets rules for this kind of puzzle, so the timer, the time penalty, the lives, the grid sizes and the difficulty schedule are this site's own choices. Outside the daily mode, every draw comes from the browser's cryptographic random numbers. Today's colours instead derives its colours from the date and works as a daily challenge, with the best result of each day kept for the last 30 days. The word grid game offers a daily grid of letters on the same idea.

The format will be familiar from phone apps such as Kuku Kube. This page is a separate browser game with no link to that app or its makers, and it needs no download.

In the timed games, Pause stops the clock and covers the colours, so nobody can study a grid while time stands still. Switching to another tab or app pauses the clock in the same way. On a keyboard, the arrow keys move between tiles and Enter or the spacebar picks one. End game stops a game early and shows its result.

Level 1 shows a 2 × 2 grid with a difference of ΔE 20.0, and from level 21 onwards every grid holds the floor of ΔE 1.0. Two things change between those points.

• The grid size grows by a row and a column every second level, from 2 × 2 to 8 × 8, reached at level 12 with 64 squares.
• The difference schedule starts at ΔE 20.0 and shrinks by 14% from one level to the next, never going below the floor of 1.0.

By level 10, for instance, the odd one hides among 49 squares on a 7 × 7 grid, with a target of ΔE 5.1. Smaller squares make the same difference harder to judge, so the last levels get tougher in two ways at once.

A screen shows each colour in whole steps for red, green and blue, the 8-bit values of a standard display. The game rounds both colours to those values and then measures the difference again, so the value shown can sit a little off the target. The result always reports that measured value.

Delta E, written ΔE, is one number for how different two colours look. The W3C's CSS Color Module Level 4 says a colour difference formula "estimates how different two colors appear", and explains that the E comes from the German Empfindung, meaning sensation. This game uses CIEDE2000, the formula in the standard ISO/CIE 11664-6:2022 from the International Commission on Illumination (CIE). A larger value means a plainer gap between the two colours, and zero means the same colour. Level 1 opens at ΔE 20.0, a deliberately clear step, while the hardest grids sit at 1.0. The smallest difference spotted in your result is the lowest value among the grids you solved in that game. It describes the colours the game sent to the screen, which is why the next sections deal with the screen itself.

The standard was written for objects that reflect or transmit light, such as printed or dyed surfaces. Displays count only where they simulate such objects, so the ΔE between two patches on a screen is an estimate, never a measurement.

To check the arithmetic, the University of Rochester publishes 34 test pairs from Sharma, Wu and Dalal (2005), each with its expected answer. The formula in this game reproduces all 34 to four decimal places.

Each colour starts as an sRGB value, the standard colour space of the web (IEC 61966-2-1). The game converts it to CIELAB, which describes a colour by its lightness and two opponent axes, green to red and blue to yellow. The conversion follows CSS Color 4 and uses the D65 white of sRGB, while CSS's own lab() function uses D50. Both tiles go through the same steps, so the comparison holds. Every standard and study named here is listed under Sources.

The game measures the gap between the colour values it sends, but the screen decides how those values look. Newcastle University's tetrachromacy researchers point out that screen colours vary considerably from one device to another.

Apple's Night Shift moves the display towards the warmer end of the colour spectrum, and Microsoft describes Night light in Windows as showing warmer colours at night. Both lower the screen's colour temperature, giving it a yellower cast. True Tone on Apple devices also adapts the colours to the light in the room.

Screen brightness plays a part as well. On a dim screen dark shades crowd together, and glare on a bright screen outdoors can wash out pale ones. Viewing a phone at an angle can shift some colours too.

None of these settings changes the ΔE in the result, which comes from the colour values rather than the light the panel gives off. They can still make the same grid look easier on one device than on another.

The Lightness only option makes the odd tile lighter or darker than the others, while its hue and chroma stay the same apart from tiny rounding shifts. Hue is the colour itself, such as red or blue, and chroma is its strength, close to what everyday speech calls saturation. The default option, Any colour change, moves hue, chroma and lightness at once, in a random direction.

Lightness carries information on its own, a point the Web Content Accessibility Guidelines make. The W3C's explanation of WCAG 2.2 success criterion 1.4.1, Use of Color, says colour must not be the only visual means of telling things apart. A significant difference in lightness, with a contrast of 3:1 or more, counts as another means.

That is why the option stays playable for many people with colour vision deficiency (colour blindness). Some hues can be hard for them to tell apart, while lighter and darker still show. The NHS describes the condition as seeing colours differently to most people and having difficulty telling colours apart.

Personal bests are kept separately for each option, so a game of Lightness only never competes with one of Any colour change. That keeps the accessibility choice from counting against anyone.

No. The game shows the smallest difference found on one screen in one session, and it diagnoses nothing about anyone's eyes.

The NHS names two main tests for colour vision deficiency. One is the Ishihara test, made of images of coloured dots; in colour arrangement tests, coloured objects are put in order of shade. It adds that colour vision testing is not part of the routine NHS eye test, but an optician can be asked for one.

Tests are available online, the NHS says, yet it is important to get checked at an optician's when there are concerns about vision. On the official nidirect site, Northern Ireland's public information service puts colour vision deficiency at about one in 12 men and one in 200 women.

The DVLA guidance on visual disorders says a driver with colour blindness may drive and need not notify the DVLA. That holds for car and motorcycle licences and for lorry and bus licences alike.

The NHS notes that the condition can rule out some careers, such as pilot or train driver. The Train Driving Licences and Certificates Regulations 2010 require normal colour vision, checked with a recognised test such as Ishihara. They also require a test of colour signals based on single colours, not on relative differences.

Relative differences are exactly what a spot the different colour game measures. Even a top score therefore says nothing about recognising a red or green signal.

About 2.3 million colours can be told apart in natural scenes, according to Linhares, Pinto and Nascimento (2008). They worked from images of 50 scenes, and found that leaving lightness aside cut the count to only about 26,000. The figure of 10 million, often repeated online, has no published origin.

Typical colour vision rests on three types of cone cell in the retina, each most sensitive to a different band of light. As nidirect explains, colour vision deficiency happens when cones are either missing or do not work properly.

Tetrachromacy means colour vision with four independent channels instead of three. Jordan, Deeb, Bosten and Mollon (2010) note that some 12% of women carry the mild inherited forms of anomalous colour vision, which can give them a fourth type of cone. Of 24 women who must carry deuteranomaly, a mild red-green form, one behaved as a tetrachromat on every test.

Newcastle University's Tetrachromacy Project explains that computer screens do not provide enough colour information, so no online test can investigate tetrachromacy. A screen mixes every colour from just three primaries, red, green and blue, so a fourth dimension of colour never leaves it.

A good score is one that beats your own earlier games in the same mode and option, since no official scale exists for this game. The score is the number of levels cleared.

Anyone still playing at level 21 has reached the floor of ΔE 1.0, which sits close to published perceptibility thresholds. Paravina and colleagues (2015) asked 175 observers to compare dental ceramic samples in a viewing booth. The perceptibility threshold, where half the observers saw a difference, was 0.8, and the acceptability threshold 1.8.

Morovic and Morovic (2023) found a threshold in the region of 0.5 to 0.8 in an online pilot with colour gradients. Neither study sets a single perceptibility threshold for every colour and screen, and a floor of 1.0, just above both, is the site's own choice.

Your personal best is kept for each mode and option. A new game beats it with more levels, then with a smaller difference spotted, then with fewer mistakes. The last ten games are listed under the board, and the daily mode shows your best for the current day.

For a score that rests on the speed of the hand rather than the eye, the click speed test counts clicks per second.