What is a market regime? The weather system your trades live in

The idea in one paragraph

A market regime is a persistent state of market behaviour — a stretch of weeks or months in which volatility, correlations and the character of returns stay recognisably similar, before switching, often abruptly, to something else. Calm uptrends where dips get bought. Sideways churn with no memory. High-volatility panics where everything falls together. The statistics of your P&L are not drawn from one distribution; they are drawn from whichever of these states the market currently occupies.

This is not a metaphor. It is one of the most robust empirical findings in finance: equity returns show volatility clustering (turbulent days bunch together), and the calm and turbulent clusters have different average returns, different tail risk and different correlation structure. Modelling markets as a small number of switching states — the idea behind Hamilton's regime-switching models from 1989 onward — consistently fits the data better than any single-distribution story.

Why it matters more than your entry signal

Almost every strategy is implicitly a bet on a regime. Trend-following earns in persistent regimes and bleeds in churn. Mean-reversion does the opposite. Selling volatility is wonderful until the state flips, and then it is ruinous. The uncomfortable truth is that a strategy's headline backtest is mostly a statement about which regimes dominated the test window.

The question "does this strategy work?" is usually the wrong question. The right one is "in which regime does this strategy work — and which regime am I in now?"

That second question is answerable. Volatility is measurable, persistence is measurable, and the switch points — while never predictable to the day — leave statistical fingerprints. That is all a regime model is: an honest, mechanical answer to "what kind of market is this?", updated daily, with a probability attached.

What a regime is not

• Not a forecast. Knowing you are in a bear regime doesn't tell you tomorrow's return; it tells you tomorrow's return is being drawn from the wide, ugly distribution rather than the narrow, friendly one. That changes position sizing, not direction calls.
• Not a chart pattern. Regimes are statistical states, defined by measured volatility and return character — not by anything an eye finds in candlesticks.
• Not permanent. The average bull regime in our 20-market sample persists for months. Persistence is what makes the concept useful — a state that flipped daily would be noise — but every regime ends, which is why the probability matters more than the label.

How to actually use one

Treat the regime as the environment, and your process as the plant that grows in it. Size positions to the regime's volatility, not to conviction. Expect trend entries to fail more in neutral churn. Read a rising bear probability as a rising cost of being wrong. And when the regime and your thesis disagree, at minimum know that they disagree — the regime read is one input, not an oracle.

Everything on this site hangs off that one concept: the HMM board estimates the state probabilities, the Compass maps which macro weather the market is pricing, and the change log records every flip we have ever published, dated, so you can audit how the calls aged.