In the Portfolio Lab, enter up to five holdings and their weights as Mix A. Add a second or third mix to compare. The Lab shows growth, the worst fall, how long each mix sat underwater and how it did against two benchmarks you choose.
Save it under My portfolios. Saved mixes are private to your account.
In Signal Backtest, pick your mix under Start from a Portfolio Lab mix. Each holding becomes a sleeve with its own rule for when it steps aside into a defensive market.
Choose which house readings decide that for each sleeve: the macro Risk Matrix, the liquidity tide and the momentum (VAMS) reading. The record sets the result beside just holding the same mix.
Macro & Liquidity shows today's readings and what followed similar readings in the past. The Asset Dashboard shows each market's reading.
Charting draws any market, or a saved mix, with the readings on it.
Is any of this a recommendation?
No. Every chart is a record of what happened, using only information that was available at the time. The tools do what you set them to do and show the outcome, including the windows where a rule did worse than holding. Nothing here considers your circumstances, and nothing here says what to buy, sell or hold.
What do the house readings measure?
The Risk Matrix reads what 40 markets worldwide are collectively signalling about growth and inflation. It lands in one of four conditions: risk on or risk off, each with an inflationary or disinflationary lean.
The liquidity reading scores how much money is available to chase assets worldwide, 0 to 100, and names it as one of four tides.
VAMS is a momentum reading for each market, from −2 to +2. The construction of all three is proprietary.
Why does a rule that worked on one market lose on another?
The same reading arrives at a different point in each market's path. A rule that stepped aside just before one market fell can step aside in the middle of another market's climb. On some markets the house rules beat holding and on others they lost, so the record on one market is one draw. The per-sleeve table under the backtest shows each sleeve's rule against holding that market alone.
Why did my backtest start later than the window I picked?
A backtest can only start once every market in it has history, so the youngest market sets the start. Treasury bills (BILS), the default defensive market, begin in September 2020. Choose SHY as the defensive market for a record that runs back more than twenty years.
Price or total return?
Price returns leave out dividends and interest, which understates income-paying markets such as bonds and credit by several points a year. Signal Backtest and Market Returns have a Total return switch for every market with a dividend record. Crypto, currencies and gold pay none, so they read the same either way.
When does a rule act on a reading?
Each decision uses the prior day's close and takes effect the next trading day, so no result uses information that arrived later. Costs are off by default. Set a cost per switch to see how often a rule traded and what that would have cost.
Where are my saved portfolios?
In your account, visible only to you. Lab mixes appear in Signal Backtest under Start from a Portfolio Lab mix and on Charting under Your portfolios. Backtests and benchmarks saved in Signal Backtest stay there.
Why is today's reading sometimes dated the day before?
The readings update after the US close. If a day's update does not arrive, the page shows the last reading with its date rather than guess.
Each bar is the share of days in that combination that a 10%+ fall followed within three months; the dashed mark on every bar is the ordinary-day rate of 23%. Figures are historical frequencies over 2007–2026, not probabilities and not predictions. Average three-month returns were positive in all four combinations: what separates them is the size of the falls along the way, not where markets finished.
| Category | Asset | Macro | Liquidity | Mean reversion | Momentum | Composite | 3M after this state |
|---|
| Market | Window | Return | Annualised | Max drawdown | Volatility p.a. |
|---|
Each dot is one rolling three-year window of US stocks against 10-year bonds. Once trailing inflation passes about 2.5%, the average window sits above zero: stocks and bonds fall together.
What bonds returned in each crisis
Real return on the 10-year bond across each crisis’s equity peak-to-trough window. Select a card above to isolate one shock type.
Every episode as a table
| Episode | Shock type | Regime | Correlation | Equity window | Equity fall | Bond real return |
|---|
How this is measured
The term premium and the stock-bond correlation
The rolling three-year stock-bond correlation (left axis) against the 10-year term premium, the extra yield bondholders take over expected short rates (right axis), monthly since 1961.
How this is measured
CPI prices a basket of goods, not the currency they are priced in. Divide a market by the money supply instead and the answer changes. Pick a market, a yardstick and a start.
How this is measured
The dollar since 1971, against ten yardsticks
How many times each has multiplied since the US closed the gold window in August 1971, on a log scale. Wages are included; the popular version of this chart leaves them out. Hover a bar for what the dollar lost against it.
How this is measured
How fast AI capability compounds, what the build-out behind it costs, and how much of that bill the builders’ own cash flow still covers.
The length of task AI can finish keeps doubling
Each point is a model: the human time of the tasks it completes half the time. Log scale, with METR’s own confidence intervals. Filled points were the best measured at release.
How this is measured
What it costs, and what pays for it
Quarterly capex by Microsoft, Alphabet, Amazon and Meta (bars) against their combined operating cash flow (line). Where the bars meet the line, current operations no longer cover the spending.
How this is measured
Capital and labour, since 1990
The same build-out in the national accounts, indexed to 100 in January 1990 on a log scale: equal slopes mean equal growth rates.
How this is measured
Savings are stored labour, spent later on other people’s labour: care, food, housing. So measure them in hours of work, not in dollars.
What things cost in hours of work
A price divided by the average hourly wage of production and nonsupervisory workers: how long an ordinary worker laboured to buy one.
How this is measured
The leaky bucket, from every start year
Hours of work commanded at the end for each hour of pay saved monthly from that start year to the latest month. Log scale; 1 is break-even. All four stores shown.
Run your own window
Store hours of your pay each month across a past window and see what they command at the end. Change any field and it recalculates.
How this is calculated
Debt to GDP holds still only when nominal growth clears a bar set by the debt, its interest rate and the budget before interest. It is an identity: it says what growth would have to be, never what it will be.
Three line items against all federal receipts
Social benefits, defence and interest, stacked as a share of every dollar the federal government collects. Above 100%, those three alone use up all receipts.
How this is measured
Interest against defence
Federal interest on the debt and national defence spending, national accounts basis, US$ billions a year.
How this is measured
Who holds the debt
Debt held by the public, split between foreign holders, the Federal Reserve and everyone else, quarterly since 1970.
How this is measured
The refinancing gap
The average rate the Treasury pays on its marketable debt against the 5-year yield, monthly since 2001. Red where the market rate sits above the average, so debt that matures is refinanced at a higher rate; green where it sits below.
How this is measured
The growth the arithmetic needs, against the growth delivered
Dark line: the nominal growth that would have held debt to GDP flat that quarter. Blue: growth delivered. Green where it cleared the bar, red where it fell short.
How this is measured
The tailwind that has already reversed
The effective interest rate on the debt against the 10-year yield, with the ageing that drives the benefits line (right axis).
How this is measured
Run the arithmetic yourself
Set three inputs and the identity does the rest: the same calculation as the growth chart, applied to numbers you choose rather than numbers that happened.