You wake up and measure your risk how? What do you do? What is Wall Street’s typical plan for measuring risk? Trend followers measure risk every day, but what about others?

This question is not abstract. The answer, or the failure to have one, has destroyed some of the most sophisticated trading operations in history. Long Term Capital Management employed Nobel Prize winners, built mathematical models of extraordinary complexity, and still collapsed spectacularly in 1998. The failure was not a lack of intelligence. It was a misunderstanding of what risk actually is and how it should be measured.

David Loeper on Risk Measurement

David Loeper of Financeware offers insight on the concept of risk. However, while we find he presents good food for thought and some very good concepts and lessons, he misses the mark on a key point. Here is an excerpt from his writings:

Getting the most return out of a portfolio for the risk being taken is extremely helpful. The way we measure this risk really is not all that material to most statisticians. Whether it is standard deviation, maximum loss or some other risk measure is not really all that important. A good mathematician can conceptualize these into his or her realm. The generally accepted measure is standard deviation. This is not a particularly difficult concept to grasp, as what it measures is the extent and frequency that individual returns will vary from the average of the returns.

We disagree with this excerpt. Measuring risk is indeed important. Unfortunately, standard deviation measures volatility, not risk. Standard deviation does not consider the ordering of returns — this is a huge point. This kind of thinking sounds similar to what drove Long Term Capital Management.

Why the Distinction Between Volatility and Risk Matters

The difference between volatility and risk is not a technical quibble. It is the distinction that separates traders who survive from those who do not. Standard deviation measures how much returns vary around their average. It treats a 10% gain followed by a 10% loss as equivalent to a 10% loss followed by a 10% gain. Mathematically, the standard deviation is identical. In practice, the ordering of those returns determines whether a trader stays solvent.

A sequence of losses early in a trading program, before sufficient profits have been accumulated, can be lethal. The same sequence of losses occurring after years of compounded gains is survivable. Standard deviation captures none of this. It summarizes the distribution of returns as if the sequence were irrelevant, which for a live trading operation it is not. A trader cannot average their way through a catastrophic early drawdown. They run out of capital before the average asserts itself.

Long Term Capital Management’s models assumed that historical correlations between assets would hold under stress conditions. They did not. When the 1998 Russian debt crisis hit, correlations between markets moved toward one. Assets that were supposed to offset each other’s risk moved in the same direction simultaneously. The standard deviation-based risk models said the portfolio was safe. The real world said otherwise. The fund lost $4.6 billion in less than four months and required a Federal Reserve-orchestrated bailout to prevent broader market contagion.

Trend followers measure risk differently. Rather than relying on statistical summaries of historical return distributions, a trend following system measures risk directly: how much of current equity is at risk on each position, what the current market volatility is, and how large a position can be held while keeping any single trade’s potential loss within predefined bounds. That measurement is done fresh every day, based on current conditions, not historical averages. The risk management framework is not a model of what markets should do. It is a daily response to what markets are actually doing right now.

The Ordering of Returns Problem

Loeper’s statement that the way risk is measured “is not all that material” is precisely the kind of thinking that leads to LTCM-style failures. When the measurement tool is insensitive to the sequence of outcomes, it cannot alert a trader to the real danger: that a string of losses at the wrong time, under the wrong conditions, can end a trading operation regardless of what the long-run average return looks like.

Trend following’s approach to this problem is structural. Position sizes are calculated relative to current equity at all times. When equity falls, position sizes shrink automatically. When equity grows, positions can grow proportionally. The system never allows a losing streak to compound into a position size that exceeds what current equity can support. This is the practical solution to the ordering-of-returns problem that standard deviation cannot address: keep risk proportional to current capital, not to some historical average of what capital has been. For the specific mechanics, see the TurtleTrader rules on position sizing and the full drawdown management framework.

More on risk and volatility: 1. More on risk and volatility: 2.

More on Long Term Capital Management from The Cato Institute.

Frequently Asked Questions

What is the difference between volatility and risk in trading?

Volatility measures how much returns vary around their average, typically expressed as standard deviation. Risk, in the context of a live trading operation, is about the probability and magnitude of loss that could end or permanently impair the trading program. Standard deviation is insensitive to the sequence of returns, which determines whether a trader survives a losing streak. A proper risk measurement framework accounts for current equity, current market volatility, and the size of each position relative to both.

Why did Long Term Capital Management fail if it had Nobel Prize-winning mathematicians?

Because the mathematical models assumed historical correlations between assets would hold under stress. When the 1998 Russian debt crisis hit, correlations moved toward one simultaneously across markets that were supposed to offset each other’s risk. The standard deviation-based risk models said the portfolio was safe. The real world produced a $4.6 billion loss in under four months. Intelligence and complexity could not compensate for a fundamentally flawed approach to measuring risk.

Why does the ordering of returns matter so much?

Because a trader cannot average their way through early losses. A sequence of large losses before sufficient profits have been accumulated can exhaust capital before the long-run average return has a chance to assert itself. Standard deviation treats all sequences with the same variance as equivalent. A trading operation that suffers catastrophic losses early does not have the luxury of waiting for the average to improve.

How do trend followers measure risk differently?

By measuring it fresh every day based on current conditions rather than historical averages. Position sizes are calculated relative to current equity and current market volatility. When equity falls, positions shrink. When volatility rises, positions shrink. The system keeps risk proportional to what current capital can support, not to what a historical statistical model says it should be able to tolerate.

Is standard deviation useless as a risk measure?

It is useful as one input among several but dangerous as the primary or only risk measure. Its failure to account for the sequence of returns and its assumption that historical distributions predict future behavior make it inadequate as a standalone framework, particularly under stress conditions when correlations and distributions change in exactly the ways the model did not anticipate.

Trend Following Systems
Want to learn more and start trading trend following systems? Start here.