How sharp should the pivot be?

Diverging colormaps pivot around a center — but how abruptly? A scroll-driven look at pivot sharpness, a county map you can sharpen by hand, why an asymmetric domain still wants a symmetric scale…

Summary

Every diverging colormap pivots around a center — but designers rarely talk about how abruptly it does so. That single choice decides whether a near-zero value reads as quiet grey or as committed color, and it can either clarify a threshold or manufacture a signal out of noise. Scroll to sharpen the pivot, drag the slider on a county map, keep both arms on one scale even when the data is lopsided, then compare how common diverging maps actually bend.

A diverging map pivots around a center.

It builds on a rule from the anatomy note: lightness stays monotonic on each arm. Here we ask a different question — not whether the arms are well-formed, but how abruptly they meet.

The pivot is the zero-crossing.

It's the single point where the map is most neutral — lowest chroma, peak (or trough) lightness — and where the encoding flips sign. Everything interesting about a diverging map happens in how it behaves right around this point.

Anchor it to something that means zero: a baseline, a threshold, a reference period — never the data median.

How fast does color leave neutral?

Sharpness is the width of the near-neutral band around the pivot — equivalently, how steeply lightness and chroma change as you step off center.

A sharp pivot has almost no plateau: the tiniest deviation already carries visible color. A soft pivot holds a wide grey zone, so values near zero look basically the same.

Sharp: a crisp sign boundary.

Collapse the neutral band and the zero-crossing becomes a hard visual edge — cool one side, warm the other, almost nothing between. The eye instantly reads which side of zero every cell is on.

Exactly what you want when the sign is the story: gain vs. loss, above vs. below threshold.

Soft: small deviations stay quiet.

Widen the neutral band and values near zero fade toward grey. Only substantial deviations earn saturated color, so the map foregrounds magnitude and lets the noisy middle recede.

What you want when near-zero wobble is just noise you'd rather not dramatize.

Sharp makes a threshold. It can also invent one.

On the county map below, a sharp pivot paints counties sitting just off zero a committed cool or warm. If that wobble is noise, the map draws a boundary the data does not have.

A soft pivot lets those counties stay grey — and can hide a small, real change that sits in the same band. The right setting depends on whether near-zero means nothing happened, or just barely happened.

Pick sharpness to fit the question.

Sharp — when crossing the reference is the headline: temperature anomalies, profit and loss, election margins, pass/fail thresholds.

Soft — when near-zero is noise to suppress: correlation matrices, dense difference fields, anything where only the large excursions should speak.

The county map below is one ramp and a sharpness slider. Then a quieter rule: keep both arms on one scale even when the data is lopsided. After that, the gallery ranks real diverging maps by how smoothly they turn at the pivot.

Change in alcohol-related deaths, by county.

Each county's relative change in deaths from alcohol use disorders, 1980 → 2014 (IHME data, after Mike Bostock). The pivot is anchored at no change: cool means the rate fell, warm means it rose. Drag the sharpness and watch counties near zero either settle into grey (soft) or snap to a side (sharp) — while the genuine clusters of rise and fall stay legible throughout.

3,142 counties, recolored live through the same map — only the sharpness of the pivot changes. Data: Institute for Health Metrics and Evaluation, via Bostock's “Methods of comparison, compared.” See also the anatomy note and the gradient diagnostics.

The same unit on both sides.

The county map above already keeps both arms on one scale. The legend is symmetric around no change, whether or not more counties rose than fell. That is the rule, not a convenience. A diverging map is a promise: equal distances from the pivot look equally colored. The promise only holds if both arms share a scale.

When the data itself is lopsided — say [−2, +10], pivot at 0 — a convenient default stretches each arm to the data min and max. Then −2 looks as extreme as +10, and +2 looks quieter than −2. The picture invents a symmetry the numbers do not have.

Two-slope scaling (each arm independently min-maxed) uses every color and maxes local contrast. It is also the option you probably never want. It is perceptually misleading. The anatomy note puts the same rule as a one-liner; this is why that one-liner exists.

Is the question how far from the pivot, and which way?

Center the range. Same unit on both arms.

Two-slope maxes contrast. It also lies about magnitude. You probably never want this.

How real maps bend at the pivot.

The hairline is the pivot. Full colormap is the ramp as published, endpoints and all. Around the pivot crops every map to the same narrow span and stretches it, which is why that view looks pale: you are looking at the near-neutral middle, not the saturated ends. A dark center (Berlin, Lisbon, Ice & fire) is a lightness choice. It is not, by itself, a sharp one. Arrow keys move the selection.

Nothing in this range. Widen the smoothness brush or clear the search.

Lightness and chroma appear once the stops are measured.

Related resources

Citation

colorcurio.us. “How sharp should the pivot be?.” Published 2026-05-30. https://colorcurio.us/notes/diverging-pivot-sharpness