TurtleTrader comment: Icebergs, Titanic and trend following are related? Bright men like Kent Osband have recently injected insight into the “risk” debate that all traders should ponder:

The single most widely used measure of financial risk is the volatility, or annualized standard deviation. It’s an attractive measure because the volatility of a portfolio depends only on the volatilities of the individual assets, their weights, and their correlations…[But] every time the normal distribution is used for portfolio analysis even when the constituent assets are clearly non-normal, the implicit answer is “yes, the approximation is relatively good.” Indeed, the implicit answer is “yes” virtually any time an analyst gauges portfolio risks simply by the mean and volatility, for with most other distributions the mean and volatility would not suffice to characterize risks.

TurtleTrader comment: Bottom line? Know your risk and we don’t mean standard deviation!

Even in large portfolios, high standard-deviation events tend to occur far more often than a normal distribution suggests. This seems especially true on the downside, which will call this variable excess risk “iceberg risk”, because it is mostly hidden from view but threatens major damage. It might also be called “Noah risk”, after the proverbial flood that drowned the world…Also, in the Biblical story, the world was amply warned about the flood but refused to listen. In contrast, icebergs reflect a type of risk that people look for but may not see.

TurtleTrader comment: From the get go you must be prepared for risks you can’t see until they happen. When they hit, you have a plan.

How I envied my colleagues specializing in Latin America. Compared with the post-Soviet bloc, a country like Mexico had much better data, a much more stable and experienced political regime, and a much longer relevant debt-servicing record. Based on extensive consultations with the IMF, with major investors, and with Mexican government officials, my guru colleagues confidently predicted the Mexican peso would hold to its announced exchange rate band against the US dollar. They were right too. They were right down to the date that they were proved dead wrong. The Mexican iceberg was a revelation for me. Wow, I thought, some of these gurus are just as ignorant as me. More productively, I realized that the markets I specialized in were bound to move largely on perceptions and changes in perceptions, with only occasional wake-up calls from reality. Ever since, I have focused on better understanding the role of ignorance in financial markets, both from a theoretical and a practical perspective. Call it an odyssey in ignorance.

TurtleTrader comment: Osband is driving right toward trend followers. Trend followers seem to be the true exception on Wall Street. They are cognizant of “iceberg” and “Noah” risk every day. If you know that markets “move largely on perceptions and changes in perceptions, with only occasional wake-up calls from reality,” then what other strategy choice beyond trend following exists?

What Iceberg Risk Means for Systematic Traders

Osband’s iceberg/Noah distinction is worth holding carefully. Noah risk is the risk that is known and warned about but ignored. The iceberg is the risk that is actively looked for but not found until it surfaces. Both are forms of tail risk, but they have different implications for how traders should prepare.

LTCM was Noah risk. The possibility that convergence strategies could face a crisis of simultaneous liquidity withdrawal was theoretically known. The models assumed it would not happen within the relevant time horizon. The assumption was wrong. The management chose to proceed. The warning was there.

The Mexican peso crisis of 1994 was iceberg risk in Osband’s framing. The analysts who monitored Mexico had better data, better political stability, and better expert consensus than most emerging markets. The consensus was wrong not because anyone was hiding the risk deliberately but because the relevant risk was genuinely not visible in the data that was being monitored. The exchange rate band that everyone consulted agreed would hold did not hold. The iceberg was there but could not be seen from the surface.

Trend following’s structural response to both types of risk is identical: do not depend on models that assume risk is normally distributed. A trailing stop defined at 2N from entry does not assume that the maximum loss on the position is bounded by the standard deviation of recent price moves. It assumes that price can move any amount before the stop fires, and the stop fires when price reaches the defined level regardless of how many standard deviations that represents. If the position opens limit-down 15 standard deviations on an iceberg news event, the stop fires and the loss is taken. The model did not need to anticipate the event. The rule only needs to fire when the event produces the price movement.

The Mexican peso observation, that markets move largely on perceptions and changes in perceptions with only occasional wake-up calls from reality, is the behavioral finance description of how trends work. Perceptions of a stable peso support the peso’s value. When the perception changes, the peso falls sharply. The trend follower who is positioned short the peso when the perception changes captures the move. They did not anticipate the perception change. They reacted to the price movement that the perception change produced. That is the only strategy choice that handles iceberg risk correctly: react to what actually happens rather than predict what should happen based on the available data.

Frequently Asked Questions

What is the difference between iceberg risk and Noah risk?

Noah risk is tail risk that is known and warned about but ignored or discounted. The Biblical flood was predicted and the warning was refused. LTCM’s liquidity risk was theoretically known but assumed to be practically negligible. Iceberg risk is tail risk that analysts look for but cannot see in available data. The Mexican peso crisis appeared stable by every conventional metric until it was not. Trend following’s response to both is identical: define exits that fire regardless of what the model predicts, because the model may have missed either the Noah or the iceberg.

Why is standard deviation an inadequate measure of financial risk?

Because financial returns are not normally distributed. They have fat tails: large adverse events occur far more frequently than a normal distribution predicts. Standard deviation characterizes risk adequately only for normally distributed outcomes. When the distribution is non-normal, the mean and standard deviation are insufficient to characterize the tail risk that will determine whether the strategy survives a crisis. Iceberg risk is precisely the tail risk that standard deviation does not capture because it does not appear in the historical variance that standard deviation measures.

How does trend following handle risks that models cannot anticipate?

By defining exits based on price movement rather than on model predictions. A trailing stop at 2N from entry does not require anticipating the iceberg. It requires only that the position be closed when price falls to the defined level, whatever the cause. An iceberg event that moves price 15 standard deviations below entry fires the stop at 2N. The loss is taken. The position is closed. The model did not need to predict the iceberg. The rule only needed to fire when the iceberg produced its price movement.

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