A split complementary scheme takes one color, finds its opposite on the wheel, then throws that opposite away and uses the two colors either side of it instead. Base plus 150 degrees, base plus 210 degrees. Three colors, arranged in a narrow isosceles triangle with the point aimed at your base.
The pitch is always the same. You get the punch of a complementary pair without the head-on collision, and a third color for free. It is the scheme every design course recommends to beginners, on the grounds that it is hard to break.
I ran the geometry through the color-difference formula this site uses for scoring, sweeping a saturated split complementary set through 120 rotations of the wheel. The pitch holds up on one axis and collapses on another. The two legs of the triangle really do keep almost all the contrast of a true complementary pair, losing about ten percent. But the third side, the one connecting the two split colors to each other, averages less than half the length of the legs, and in a quarter of all rotations the two colors you chose as your accents are near enough the same color. The numbers are below.
What a split complementary scheme is
Start with any hue. Look directly across the wheel, 180 degrees away. That is its complement. Now step 30 degrees back from that point in each direction. Those two hues, plus your original, are a split complementary set.
Blue at 240 degrees gives you orange at 30 and chartreuse at 90 rather than the yellow at 60 you would have got from the straight complement. Red at 0 gives you spring green at 150 and azure at 210 instead of cyan at 180.
The 30 degree step is convention rather than law. Older texts draw it on a twelve-part wheel, where one step is 30 degrees by construction, so the split lands on the two hues flanking the complement. Some guides widen it. Nobody explains why 30 rather than 20 or 45, which turns out to matter a great deal, and I come back to it below.
The scheme is sometimes written as a compromise between two others. It has one strong opposition, like a complementary pair, and three members, like a triad. That framing is closer to the truth than most people realise, in a way that is only obvious once you measure it.
Split complementary examples with hex codes
These are the sets people are looking for. Hue angles are on the digital wheel, the one behind every HSL and HSV color picker, at full saturation and brightness so you can paste them straight into a design tool and adjust from there. If the notation is new to you, how to read hex color codes takes about two minutes.
- Blue base. #0000ff with #ff8000 and #80ff00. The textbook example. Blue against orange and yellow-green, avoiding the pure yellow that would have been its complement. Warm accents on a deep cold field, which is roughly every science fiction poster ever printed.
- Red base. #ff0000 with #00ff80 and #0080ff. Red against spring green and azure. The most usable of the classic sets in my view, because the two accents are genuinely different from each other rather than two shades of the same idea.
- Yellow base. #ffff00 with #0080ff and #8000ff. Yellow against azure and violet. Very high drama, and the one that suffers worst from the problem described in the next section.
- Green base. #00ff00 with #8000ff and #ff0080. Green against violet and rose. Reads as synthetic and deliberately so, which is why it turns up in club flyers and almost nowhere else.
- Orange base. #ff8000 with #00ffff and #0000ff. Orange against cyan and blue. The teal and orange grade that took over film posters around 2005 is this set with the saturation pulled down.
- Violet base. #8000ff with #ffff00 and #00ff00. Violet against yellow and green. Heavy, and the least balanced of the six once you look at the numbers.
Full saturation is a starting point, not a requirement. Tint, shade or mute any of these and the scheme still holds, because the hue angles have not moved. A navy with a rust and an olive is the blue set several steps quieter, and it is a far more likely thing to ship.
One caveat on the hex codes. These sit on the digital wheel. If you are mixing pigment rather than picking pixels you are on the artist wheel, where the hues are in different places, and a set named the same way will not produce the same colors. That split runs through every scheme on this site and I set it out in full in the triadic colors piece.
The bridge is the weak side
Here is the measurement that changed how I think about this scheme.
The tool is CIEDE2000, which reduces a pair of colors to a single number for how different they look to a human. About 1.0 is the smallest difference a trained observer can reliably spot. Anything past 50 is unmistakably two separate colors. I walked a saturated split complementary set through 120 rotations of the wheel, three degrees at a time, and measured all three sides of each triangle.
Averaged across every rotation:
- Base to first split, one leg of the triangle: ΔE 81
- Base to second split, the other leg: ΔE 81
- Split to split, the bridge between the two accents: ΔE 41
The legs are twice the bridge. This is not a set of three colors that relate to each other equally. It is one strong opposition with a pair of near-neighbors hanging off the far end, and the two accents are close enough that many viewers will read them as variations of one color rather than two.
The averages hide worse cases. The bridge falls below ΔE 50 in 81 of the 120 rotations, which is 68 percent, and below ΔE 30 in 31 of them. So in roughly a quarter of all split complementary schemes, the two accent colors are separated by less than the gap you would call clearly distinct.
The yellow set from the list above is one of them. Yellow #ffff00 with azure #0080ff and violet #8000ff has legs of ΔE 78 and 99, and a bridge of ΔE 28. The violet base set is worse still, at ΔE 23 between its yellow and its green. Both of those pairs are inside the range I would call analogous rather than contrasting, and analogous pairs average around ΔE 22 at 30 degrees apart.
The worst rotation on the wheel is a magenta base at 291 degrees, with splits of chartreuse #a6ff00 and spring green #00ff59. Legs of 105 and 110. Bridge of 12. That palette has two colors in it, one of which happens to be printed twice.
What splitting actually costs you
The good news is real, and it is worth stating precisely because most guides assert it without a number.
A true complementary pair, 180 degrees apart, averages ΔE 92 across the wheel. A split complementary leg averages ΔE 81. Splitting the complement costs you about ten percent of the available contrast, and gives you a third color in exchange.
That is a good trade and it is the honest case for the scheme. Ten percent is not nothing, but it is nowhere near enough to break the opposition. Your base still fights both accents at close to full strength. The vibration people complain about in complementary pairs, that shimmering edge where two opposites of similar lightness meet, is genuinely reduced, because neither accent is your exact opposite and so neither one lands on the retinal geometry that causes the effect.
The cost is not evenly distributed. In some rotations splitting actually increases the contrast on one leg, by up to ΔE 11, because the wheel is not perceptually uniform and stepping 30 degrees off the complement can move you toward a region of larger differences rather than away from one. Orange at 30 degrees is an example. Its true complement, azure #0080ff, sits at only ΔE 56, an unusually weak complementary pair, while the split to blue #0000ff measures 68. You get more contrast from the split than from the complement it replaced.
The scheme is a triad with the tuning knob turned
I swept the split angle from 10 degrees to 90 to see which value produces the most even triangle, measuring the gap between the longest and shortest side and averaging that over all 120 rotations. The answer is unambiguous and, once you see it, faintly funny.
- Split of ±10°: legs ΔE 90, bridge ΔE 15, average spread 80
- Split of ±30°, the standard: legs ΔE 81, bridge ΔE 41, average spread 54
- Split of ±45°: legs ΔE 75, bridge ΔE 56, average spread 40
- Split of ±60°: legs ΔE 69, bridge ΔE 69, average spread 36
- Split of ±90°: legs ΔE 56, bridge ΔE 92, average spread 53
The most balanced split angle is exactly 60 degrees. At 60 degrees the two splits sit at base plus 120 and base plus 240, which is a triad. The function has its minimum precisely where the scheme stops being split complementary and becomes the other scheme entirely.
So a split complementary set is a triad that has been deliberately detuned, and the standard 30 degree split is halfway to fully detuned. Every degree you pull the two accents in from 60 buys base dominance and pays for it in accent separation. That is a legitimate thing to want. It is just not what the scheme is usually sold as. The common description, balanced but less harsh than complementary, has it backwards: the scheme is less balanced than a triad, by ΔE 54 of average spread against the triad's 36, and it is less harsh precisely because it is less balanced.
This also gives you a dial rather than a rule. If your two accents are reading as the same color, widen the split. Going from ±30 to ±45 raises the bridge from ΔE 41 to 56, which crosses the threshold into clearly distinct, and costs the legs only 6 units. On the measurements that is the best-value adjustment available anywhere in this scheme, and I would now reach for ±45 by default and only tighten toward 30 when I specifically want the accents to read as a set.
Lightness, which the scheme does not fix either
Hue distance is one axis. The other is lightness, measured as CIE L*, a scale from 0 for black to 100 for white built so equal steps look equal. Fully saturated hues have wildly different intrinsic L* values. Yellow #ffff00 sits at 97, nearly white. Blue #0000ff sits at 32.
Averaged over all rotations, the lightness spread across a saturated split complementary set is 44 L* points. The figure for a triad is also 44. Splitting the complement does nothing at all for the lightness problem, which makes sense once you notice that the two accents are usually on the same side of the wheel and therefore usually on the same side of the light-dark divide.
The base is the outlier. On average the base sits 33 L* points away from the mean lightness of its two accents, while the two accents sit only 19 points apart from each other. That is the actual visual structure of the scheme, stated plainly: one color at one lightness, two similar colors at another. Very close to a two-color design with a bit of variation in the second color, which is why split complementary work so often reads as clean and slightly conservative.
Munsell separated hue, value and chroma into independent axes in 1905 so that value could be chosen rather than inherited. It is still the most useful correction available. Darken the yellow toward ocher, lift the blue toward periwinkle, and you have removed the imbalance instead of rationing area to hide it.
The best and worst sets on the wheel
Ranking all 120 rotations by how close the three sides come to equal length gives a clear winner and a clear loser.
Most balanced: a vermilion base at 24 degrees, #ff6600, with turquoise #00ffe5 and deep blue #001aff. Legs of ΔE 64 and 61, bridge of 68, a spread of under 7 units. The bridge is actually the longest side, which happens in only a handful of rotations. If you want a split complementary scheme where all three colors carry equal weight, this region of the wheel is where it exists. The lightness values are 62, 90 and 34, so it still needs the usual value work.
Least balanced: the magenta base at 291 degrees already mentioned, with a spread of 98 units between its longest and shortest side. Avoid, or widen the split until the two greens separate.
The classic sets land in between. Blue base measures 68, 85 and 58, a spread of 27, and is genuinely one of the better ones, which is probably why it became the textbook example. Yellow base measures 78, 99 and 28, a spread of 71. Both are taught the same way, with the same confidence, and one of them is nearly three times more lopsided than the other.
When to use it, and when not to
Taking the measurements at face value, here is where I would and would not reach for this scheme.
- Use it when one element must clearly win. The base against two accents at ΔE 81 is a strong hierarchy that is hard to get wrong. This is the case the scheme is genuinely good at, and it is why it works so well for a single call to action on a page.
- Do not use it when the two accents need to be read as different things. At ΔE 41 average they will not reliably separate. If the two accents mean different things in your interface, a warning and an information state for example, pick a triad or widen the split.
- Check the bridge before you commit. Put your two accents next to each other, on their own, with the base nowhere in sight. If they look like one color having a bad day, the scheme has failed and no amount of area control will rescue it.
- Widen to ±45 by default. Costs almost nothing on the legs, buys a great deal on the bridge.
- Set the lightness on purpose. The scheme leaves a 44 point L* spread lying around and will not clean it up for you.
Compared against the neighbors in this family: complementary is two colors at ΔE 92, maximum tension, no room to move. Analogous is neighbors at ΔE 22, calm and prone to going flat. Triadic is three colors at ΔE 69 each, the most even option and the one that most needs area control. Split complementary is a triad detuned toward one of them. All four are the same measurement applied at different angles, which is the useful thing to hold onto when a guide presents them as four separate traditions.
Feeling the distances instead of reading them
Every number here describes something your eye already knows and cannot put into words. The quickest way to convert it into intuition is to put your eye under a little pressure and see where it fails.
Color Mixer hands you a target and asks you to build it out of red, green and blue light. It forces you onto the digital wheel, which is the wheel all the hex codes above live on, and after a few rounds the relationship between a hue angle and a visible difference stops being abstract.
Hue Sort is the more direct test for this article. You get a gradient with the middle shuffled and you put it back in order. It is a browser version of the Farnsworth-Munsell chip test, and it will tell you which arcs of the wheel you personally have resolution in. If your weak arc happens to contain both of your split accents, that bridge is even shorter for you than the ΔE says, and you should widen it further than the numbers suggest. Train your eye for color collects the drills that actually move the needle.
The short version
A split complementary scheme is a base color plus the two hues either side of its opposite, normally 30 degrees off in each direction. The classic sets are blue with orange and yellow-green, red with spring green and azure, yellow with azure and violet.
Measured, the two legs keep ΔE 81 of contrast against a complementary pair's 92, so the trade is a good one. The bridge between the two accents averages only ΔE 41 and drops under 30 in a quarter of rotations, so the two accents frequently read as one color. The most balanced split angle is 60 degrees, at which point the scheme has turned into a triad, which tells you what split complementary really is: a triad detuned to let the base win.
Then play a few rounds and find out how accurately you can place a hue from memory. Most people are worse at it than they expect, and the gap between the wheel you were taught and the wheel you can actually see is where every one of these numbers comes from.