The S&P 500 heatmap

Each tile is one company in the S&P 500, grouped by sector. Tile size represents what the company is worth; colour represents how its share price moved over the selected window. Green is up, red is down, and a deeper shade indicates a larger move.

Grouping by sector is what the map adds over a ranked table. A session where energy is uniformly green and technology uniformly red looks entirely different from one where the index moved the same amount because two large companies did all the work, and a list sorted by percentage change cannot show the difference.

Each tile links to that company's page, where the same move appears alongside the financial statements, cash flow history, dividend record and valuation multiples behind it.

A 3% move in a day is substantial. A 3% move over a year is not. A single scale applied to both would leave the one-day view almost entirely grey and the one-year view almost entirely saturated, and neither would show anything.

The scale is therefore tied to the window, reaching its deepest shade at the points below. A vivid tile always indicates a large move for the timeframe on screen.

The year-to-date view is a window that lengthens through the year, so its scale moves with it: the one-day scale on the first trading days of January, then the one-week scale, one month from mid-January, three months from late February, six months from late May and one year from mid-October.

The consequence is that colours are comparable within a view and not between views. A deep green tile on the one-day map and a deep green tile on the one-year map are not the same result, and switching the window changes what a shade means rather than only what is shown.

Tiles are sized by the square root of market capitalisation rather than by market capitalisation itself. The largest company in the index is worth several thousand times the smallest, so at true proportion a handful of megacaps would fill the screen and several hundred companies would be invisible.

The square root compresses that range while preserving the ordering: larger companies still get larger tiles. A floor keeps every constituent above about 3% of the largest tile, so the smallest remain visible and clickable.

The distortion is deliberate and it affects reading by eye. Judging what share of the index a sector represents from tile area will understate the largest companies and overstate the tail, so the map is a good way to see where movement is concentrated and a poor way to measure index weight.

The question the map answers better than any table is whether a move is broad or narrow. A sector that is uniformly coloured is responding to something that affects all of it: a rate decision, an oil price, a regulatory announcement. A sector where one large tile is coloured and the rest are grey is responding to one company.

Divergence between sectors is the other pattern worth looking for. Days when defensives are green and cyclicals are red, or the reverse, describe a change in what the market is pricing rather than a change in its level.

Both readings survive being wrong about any individual company, which is what makes them the useful ones.

It shows price, not value. A company can be deep green for a week and remain expensive relative to what it earns, or deep red and remain so.

It shows the index, not the market. The S&P 500 is large-cap US equity, so small caps and companies listed elsewhere are absent, and a session that looks calm here can be volatile in markets the index does not cover.

It shows one window at a time, and it shows price change rather than total return, so dividends are not included. Over a one-day window that is immaterial; over a year it is the difference between a flat-looking tile and a positive return.

Constituents, sectors and company sizes are rebuilt from company filings rather than copied from a data vendor, and price changes are computed from the same daily closes used across the rest of the site, so a figure here and the same figure on a company page cannot disagree.

Where a company's size cannot be established with confidence, the map shows no figure rather than an estimate.

Comparing shades between two windows, when the scale is different in each.

Reading tile area as index weight, which the square root deliberately compresses.

Watching the one-day view exclusively. Most single sessions are noise, and the three-month view frequently reads the opposite way.

Treating a uniformly red sector as a verdict on the companies in it, when a sector move is usually a response to something outside them.