The Sharpe ratio is the most-cited number in investment performance. It is also the most context-free. A single Sharpe of 1.2 over a fifteen-year sample says nothing about what the strategy did during the four worst years, or the four best, or whether the strong years and weak years cluster in ways that matter for the next investor who holds it.
This is the secret most reported Sharpe ratios are hiding. The number is an average across regimes that look nothing like each other. Strategies that perform very differently in expansions versus contractions, in low-vol versus high-vol environments, in rising-rate versus falling-rate periods, can produce the same unconditional Sharpe ratio. That ratio is not wrong. It is just compressing information that mattered.
A Concrete Example
Consider two long-only equity strategies, both with a fifteen-year Sharpe ratio of 1.0. The first delivers a Sharpe of roughly 1.4 during normal-volatility regimes and a Sharpe of roughly 0.3 during stress regimes. The second delivers a Sharpe of roughly 1.0 in both regimes. The unconditional number is identical. The risk profile is not.
An investor who holds the first strategy during a stress regime is holding something very different from the strategy the Sharpe ratio described. The number on the marketing sheet was a long-run average. The experience during the year that matters most is not the average.
Why Regimes Matter More Than Averages
Markets are not stationary. The statistical properties of returns — volatility, correlations, autocorrelation, skew — change measurably between regimes. A strategy that exploits one set of properties will work in some regimes and not others. This is not a flaw to engineer around. It is the structure of markets.
The investor's question is not "what is the long-run Sharpe of this strategy." The investor's question is "what does this strategy do in the kind of market we are about to enter, and how confident am I that we are about to enter that kind of market." Those are very different questions, and the unconditional Sharpe ratio cannot answer either of them.
An unconditional Sharpe ratio is a summary statistic for a regime that does not exist. The market is always in some specific regime. The strategy's performance is always conditional on that regime.
What Conditional Sharpe Actually Shows
When you classify historical periods into regimes — bull, bear, high-vol, low-vol, expansion, contraction, however you want to cut it — and recompute the Sharpe ratio within each regime, several things become visible that the unconditional number hid:
- Which regime is doing the heavy lifting. Many strategies that look attractive on a headline basis are carried by one regime. If that regime ends, so does the attractive performance.
- Whether the regime that hurts is rare or persistent. A strategy that earns a 1.5 Sharpe in 80% of regimes and a 0.2 Sharpe in 20% is different from one with the same numbers if the 20% regime is persistent or recurring.
- How much regime concentration risk you are carrying. A diversified-looking portfolio whose components all do well in expansions and badly in contractions is, in regime terms, still concentrated.
How We Use This
The HMM regime classification we run on the broad market is not a forecast. It is a description of which regime the market is currently most likely to be in. When we evaluate a strategy or a portfolio, we look at its performance separately within each regime — and we look hardest at the regime our model says we are currently in.
A strategy with a strong unconditional Sharpe but a weak conditional Sharpe in the current regime gets less weight, or no weight, in periods when that regime is persistent. The same strategy may earn its weight back later, when the regime changes. This is not market timing in the colloquial sense. It is statistical honesty about a fact the unconditional Sharpe was hiding: that the same strategy is, effectively, multiple different strategies depending on the environment.
One number cannot describe a strategy that behaves differently in different conditions. The first step toward better risk management is admitting that the conditions vary.
None of this is an argument against the Sharpe ratio. It is an argument against treating a single Sharpe ratio as if it described a stationary world. The investor who looks only at the headline figure is not making a wrong decision exactly. They are making a decision based on a summary statistic of a market that has not existed for any of the years they actually have to live through.
This piece discusses risk-adjusted return metrics for educational purposes. Referenced research findings describe historical outcomes and do not guarantee future results. This does not constitute investment advice. This does not create an advisory relationship.