A barometer, not a forecast

Crashboard takes forty-three measures of financial stress, folds them into two numbers, and paints the screen by where today sits against every day since 1998. This guide explains what you are looking at, and what the two numbers are not.

The two numbers

Almost everything in the app is built from two readings. They answer different questions, they are built from completely separate evidence, and they are allowed to disagree — the disagreement is often the most interesting thing on the screen.

No indicator appears in both. That matters: when one lens moves and the other does not, it is because different evidence is pointing in different directions, not because of an accounting quirk.

Each reading runs from −1 to +1, though in practice both live in a much narrower range — which is why the app prints a σ figure beside every score, saying how unusual today is against the past 28 years, rather than asking you to judge whether −0.377 is a big number.

Reading the screen

Here is the app on an ordinary day. Nothing is hidden behind a menu; the whole model is on one page. The notes below key to it, top to bottom.

The main screen A day the lenses disagree 14 October 2026

1. The first line, and the colour The whole background is one number — the Conditions Ahead reading, coloured from green to red by where it sits against the record. You can tell what the app thinks from across the room without reading a digit. Above it, on a day the two lenses disagree, a line names which of them is the lower one; on an ordinary day there is nothing large there at all, which is deliberate.
2. The two readings, side by side Each lens stands on its own vertical scale, marked once per standard deviation of its own twenty-eight years, and shows its score, how far from an ordinary day that is, and a plain-English verdict. Putting them next to each other is the point: the gap between them is as informative as either number.
3. Around the world The OFR stress index for the United States, other advanced economies and emerging markets. This is not our number: the figures are the OFR's own, with only the sign turned over so that higher reads better here as it does everywhere else, and it is free. "% Calm" says how much of the past 26 years was more stressed than today.
4. One lens at a time The switch under the world panel chooses which lens everything below it describes — the recent months, the whole record and the indicator table. The verdict and the colour above it are computed from Conditions Ahead whichever side the switch is on.
5. The recent past Six months of the lens you have chosen, each band coloured by where that day ranked against all 28 years. A washed-out stretch means the panel was split, not that nothing happened.
6. The whole record Twenty-eight years in one strip. Drag to scan, pick a date, or jump straight to the dot-com bust, 2008 or COVID with the buttons on the right.
7. What is driving it Under the charts, every indicator, what it reads in its own terms, and what that contributes to the score. These are separate on purpose: "World Uncertainty, high, −0.12" tells you both that uncertainty is elevated and that it is barely moving the needle. Faint rows are waiting on their next data release, not unimportant.
8. Now and Ahead together At the foot of the screen, the forward lens slid forward over the present one, with the market's own fall from its peak underneath. If the split works, Ahead should run in front; the slider lets you decide by how much.

Three days

The fastest way to learn the app is to visit three days. Every screen below is the same app, the same indicators and the same arithmetic. Only the date is different.

Ahead low2006-05-01

Now low2007-10-01

Today2026-10-14

The middle screen is the one worth sitting with. On 6 March 2020 the forward lens read comfortably better than the present one, which was already registering stress: policy had moved, and cheap money genuinely does read as a supportive setup. Both were true when they were taken, and the fastest drawdown in the record began within the fortnight. A single number would have had to pick one of the two and would have been wrong either way — and the forward lens, left to speak alone, was wrong here about as loudly as it can be. That is the case to keep in mind whenever it reads well.

Colour, σ, and the five verdicts

σ means "how unusual". Zero is an average day. −1σ is worse than about 84% of days in the record; −2σ is worse than about 97%. It says rare without pretending to say likely.

The verdicts are cut by rank, not by round numbers. They are the middle half and the middle ninety per cent — the same two intervals a box plot draws — so the alarm state covers one day in twenty, deliberately. Note that five bands are not fifths: a dashboard that calls one day in five a crisis has taught you to ignore it.

The history strips use the same five colours, a week at a time. Colour steps where the verdict steps, so a change of shade on a strip is a change of verdict and nothing finer. The shades move in brightness as well as hue, which is what keeps them apart for the very large number of readers who cannot separate red from green.

Understanding its output

The most useful thing to know about Crashboard is what its two numbers are not. They are not prices, not an index and not a portfolio. They measure the plumbing — credit spreads, funding costs, lending standards, employment, borrowing costs — and the plumbing does not move in step with the market.

The two readings relate to each other

Conditions Ahead measures pressure building; Conditions Now measures what has arrived. If the split works, Ahead should run in front. This chart does not try to show that. Both lines are drawn on the day each one was actually read, with nothing slid, and on those terms they sit within half a standard deviation of each other on forty-six per cent of the days drawn, crossing sixteen times. Sliding the forward line to the right would make them agree more closely — markedly so around eighteen months — and that is exactly why it is not done here: a picture slid until it looks convincing cannot be evidence for itself. Exhibit 7 of the white paper makes the lag argument properly, by running the same comparison backwards as a control and reporting each half of the record separately.

Neither one is the stock market

Here is the same chart with US equities laid underneath — how far the market had fallen from its previous high, on the right-hand scale. This is the comparison worth studying, because it separates two things that look identical on a news ticker.

A falling market is not by itself evidence that anything is wrong with the financial system — and a calm market is not evidence that nothing is. Crashboard is built to tell those two apart.

What that means when you open the app

• A red screen does not mean the market is down, and a green screen is not a buy signal. The colour is a reading of conditions, nothing else.
• The app can read calm while your portfolio falls. That is 2022, and it is not a malfunction — it is the app telling you the fall was about price, not about plumbing.
• It can also read badly after prices have already recovered, because credit and employment heal more slowly than markets do. Early 2009 looks like that.
• Conditions Ahead is not a market-timing signal. Its single lowest reading in the whole record came in June 2006. Two years later the crisis arrived — by which time the forward reading had eased back toward normal. It marks pressure, not a date.

What it will not do

• It will not tell you what to buy or sell. There are no positions, no signals and no targets, and the app has no idea what you own.
• It will not tell you the size or the timing of what is coming. The lowest single Conditions Ahead reading in the whole twenty-eight-year record came in June 2006, two years before the financial crisis. The pressure was real and it was measured. What the reading could not say was how large the event would be, or that it was still two years off.
• It cannot see what is not in the data. A pandemic, a war, a fraud — none of these are visible in credit spreads until they already are.
• Some indicators run late. Government data publishes on its own schedule; faint rows are waiting. Tap the ? beside any indicator for the date it is current through.
• Revisions happen. Macro data is revised for years after the fact, so a reading for a day in 2008 is scored with the figures as they stand today, not the ones that were published at the time. Where a number was revised heavily, the reading moved with it.
• The past is ranked with hindsight. Every reading is placed against the whole twenty-eight-year record — including the years that came after the day being shown. A day in 2008 is ranked against 2020 and 2026 as well, and it is today’s panel at today’s weights doing the ranking. So a reading you scrub back to is where that day sits in twenty-eight years of hindsight; it is not what this app would have printed at the time. Conditions Ahead is the one to watch here. Because the forward lens has drifted lower across the record, older days tend to read calmer in hindsight than they would have read live.

Data, privacy and price

Where the numbers come from

Eight United States government sources: Federal Reserve Economic Data, the CFTC, Treasury Fiscal Data, the Office of Financial Research, the FDIC, the EIA, FEMA and the USGS. All public domain. Every indicator is built from those. The equity line in Chart B above is not: it is OECD share-price data, which the OECD permits anyone to republish and asks to be cited for, as it is under the chart. This product uses FHFA data but is neither endorsed nor certified by FHFA. No licensed market data is used or republished, and the app refuses to publish a series it cannot legally carry — a rule enforced in code rather than by good intentions. Full attribution is in the app's About screen and on the home page.

The one estimated number

Every indicator is a formula over figures somebody else published, with a single exception, named here rather than left to be discovered. One of the Conditions Ahead indicators, Equity Valuation, compares the market value of US corporate equities against the size of the economy. The Federal Reserve publishes that market value once a quarter and seventy to a hundred days after the quarter it describes, so the figure spends most of its life months out of date. Between prints Crashboard estimates it from the Office of Financial Research's own daily series, by a regression refit each time a new quarter lands.

Three things about that estimate. It is replaced by the published figure the moment one arrives. It is aged as the quarterly number it stands in for rather than as a fresh reading, so it counts for less the longer the quarter has been closed. And it is an estimate of what that number is today — there is no statement in it about any day in the future, and no other number in the app is estimated at all. The white paper gives the method and how well it does against the prints it stands in for.

What the app knows about you

Nothing. No account, no sign-in, no analytics, no advertising identifier and no tracking of any kind. The app downloads one public data file and does all the arithmetic on your device. See the privacy policy.

Free and paid

Every part of the app is free for the first seven days, from the moment you open it: no card, no sign-up, nothing to tap. It is not a trial of the subscription and does not become one — nothing is charged, nothing renews, and at the end of the week the app returns to the free view. The week is counted per device, so a phone and a tablet each get their own.

Both readings are free and stay free — Conditions Now and Conditions Ahead, with the verdict and the colour computed from them — as is everything present-tense: every indicator behind Conditions Now, the global stress panel, and the whole twenty-eight-year record of it. The subscription unlocks the arithmetic behind Conditions Ahead: its indicator panel and its history, for $29.99 per year, auto-renewing until cancelled. Payment is charged to your Apple Account at confirmation; it renews unless cancelled at least 24 hours before the period ends, and you can manage or cancel it in your Apple Account settings. Cancellation and refunds are covered on the support page.