Risk Adjusted Returns Explained: Why Raw Return Is Never the Whole Answer

A risk adjusted return measures how much return a portfolio earned for the risk it took, rather than how much it earned full stop. It exists because a return figure on its own is an incomplete sentence: two portfolios can deliver identical returns while one held steady and the other put its owner through a fall deep enough to make holding it impossible.

The general shape of every such measure is the same. Put a return figure on top. Put a risk figure on the bottom. Divide. The whole family differs only in what each side means, and every argument about which measure is right is really an argument about what risk means for the decision in front of you.

What it measures

Take the numerator first. It is almost never the raw return. Most measures use an excess return: the portfolio’s return minus something the investor could have had without taking the risk in question.

• Subtracting a risk free rate, usually a short government instrument, asks how much was earned beyond simply parking the money.
• Subtracting a benchmark’s return asks how much was earned beyond the market exposure the portfolio was supposed to represent.
• Subtracting a minimum acceptable return asks how much was earned beyond what the investor needed.

Each subtraction defines what counts as “for free” and therefore what the manager should not get credit for.

Now the denominator, which is where the family branches:

Measure	Risk in the denominator	The question it answers
Sharpe ratio	Total volatility of returns	Return per unit of overall variability
Sortino ratio	Downside deviation only	Return per unit of bad variability
Calmar ratio	Maximum drawdown	Return per unit of worst realised loss
Treynor ratio	Beta against an index	Return per unit of market sensitivity
Information ratio	Tracking error against a benchmark	Active return per unit of benchmark deviation

Reading down that column is the fastest way to understand the whole subject. There is no single correct denominator, because there is no single correct definition of risk. A pension trustee worried about funding volatility, a family office worried about a drawdown deep enough to force a change of plan, and an allocator judging one sleeve inside a larger portfolio are asking three different questions, and each has a natural denominator.

How to read it

Match the measure to the decision. If the concern is whether a fall would be tolerable, maximum drawdown in the denominator speaks to that directly and volatility does not. If the holding is one component of a diversified whole, only its market-linked risk survives diversification, which points to a beta based measure. Choosing the measure first, then computing it, is the honest order. Computing several and quoting the best one is not.

Insist on identical periods, frequencies and conventions. Nearly every apparent difference between two published figures dissolves when you discover one was computed on monthly returns over three years and the other on daily returns over one. Annualisation conventions, the risk free rate used, and the benchmark chosen all shift results.

Treat short histories with suspicion. These are ratios of estimates, and both the numerator and the denominator are noisy. Over a year or two, the figure is dominated by chance. Distinguishing genuine skill from luck in a return series is statistically hard and takes far more data than most track records contain.

Read the two halves separately before reading the ratio. A ratio can improve because return rose or because measured risk fell, and those are different stories. A portfolio that shifted into less volatile holdings and one that found better ideas can print the same improvement.

Cross-check with a path measure. Volatility and downside deviation are computed from an unordered bag of returns and are blind to sequence. Drawdown is not. A portfolio can look strong on a dispersion-based ratio while having spent years underwater, so read at least one path-based statistic alongside.

Compare against a fair benchmark, not a convenient one. Any measure built on excess return over a benchmark is only as meaningful as the benchmark. This is why benchmark selection is a substantive decision rather than an administrative one.

Look at the ratio’s stability, not one print. A rolling series of the ratio over time shows whether the result came from a consistent process or from one exceptional stretch that still dominates the trailing window.

What it does not tell you

It does not tell you whether the return will continue. Every one of these measures is computed from history. Nothing in the arithmetic forecasts anything, and a strong past ratio carries no guarantee about the future.

It compresses a rich reality into one number. A single ratio cannot express the shape of a return distribution, the length of the worst underwater stretch, the concentration of the holdings, or the conditions under which the strategy works. It is a summary, and summaries discard information by design.

It can be flattered by illiquidity. Assets that are priced infrequently report smoother returns, which shrinks every volatility-based denominator and inflates every ratio built on one. The improvement is a measurement artifact rather than an economic one, and it makes comparisons across liquid and illiquid holdings unreliable.

It can be flattered by rare, large losses. Strategies that generate many small gains and an occasional severe loss will display excellent risk adjusted returns until the loss occurs, because the losing observations are not yet in the sample. No ratio computed on a period that has not contained the bad event can warn you about it.

It says nothing about why. A ratio cannot distinguish return earned from a repeatable process, return earned from one concentrated bet that happened to work, and return earned from being in the right sector during a favourable stretch. Attribution answers that, not a single number.

It does not survive a careless backtest. A ratio computed on a simulated history that excluded companies which were delisted, used financial data that was later restated, or ignored transaction costs will be higher than anything achievable in reality. Discipline about point in time data and transaction costs determines whether the ratio describes an investable result or a spreadsheet.

It ignores everything outside the return series. Governance quality, key-person dependence, regulatory exposure, capacity limits, and the honesty of the reporting are all real risks. None of them appear in a return series until after they have caused damage.

It offers no threshold for “good”. There is no universal level at which any of these ratios becomes acceptable. Levels vary by asset class, strategy, period and market regime. Anyone quoting a fixed cutoff is asserting a convention, not a fact.

The practical takeaway is unglamorous. Read at least one dispersion-based measure and one path-based measure. Read both halves of each ratio separately. Check the period and the conventions before comparing anything. And treat the whole family as a way of asking better questions about a track record, not as a scoring system that settles them.

Related reading

• Portfolio metrics explained: the hub connecting every risk and return measure.
• What is the Sharpe ratio: the most quoted member of the family, and its known weaknesses.
• Sharpe vs Sortino vs Calmar: which of the three answers which question.
• What is maximum drawdown: the path-based risk measure every dispersion ratio misses.
• Volatility and standard deviation explained: the denominator most of these ratios start from.

This article is educational. Altys Labs is not a registered research analyst or investment adviser, and nothing here is investment advice or a recommendation to buy, sell, or hold any security.