A safety limit breached twice as often as promised. A quiet bond market that authorised sixteen times leverage.
The Evidence Base:
Over 640,000 daily observations across 68 global futures markets, spanning more than four decades, from September 1984 to July 2026.
One consistent method throughout.
Standard deviation is not an academic curiosity. It is wired into the machinery that decides how much a bank may borrow, which manager gets fired, what an option is worth, and how much capital a regulator demands.
We took the six places it does the most work and tested each against the same 68 markets.
1. The safety limit that is not safe
Banks measure danger with something called Value at Risk. In plain terms it is a line drawn on the floor. We do not expect to lose more than this, except on one day in a hundred.
The line is drawn using standard deviation.
What should we see if the model is right? Losses beyond the line on 1% of days.
What we found: losses beyond the line on 1.81% of days. Close to double. And 67 of the 68 markets break the limit more often than the model allows.
The frequency is the smaller problem. When the limit does break, the average loss is more than a third larger than the limit itself, and the worst breach in four decades was more than eleven times the limit.
The line tells you where the floor is. It has nothing to say about how far below the floor the basement goes.
2. The leverage trap
This is the one that does the killing.
Most professional investors size their positions using volatility. Quiet market, hold more. Wild market, hold less. The intention is sensible enough: keep the amount of risk roughly constant.
But look at the arithmetic. Position size is set by dividing a target by recent volatility. When volatility falls you are dividing by a smaller number, so the position gets bigger. Automatically. Nobody makes a decision.
Across our universe this machinery runs more than triple the leverage in quiet markets compared to stressed ones. Precisely, 3.1 times.
The defence is real, and we already knew it
Quiet markets really are safer. We did not need to be persuaded of this. Our earlier Fractals of Finance work had already established it directly: the memory in volatility is deep and universal, with a Hurst exponent on absolute returns averaging 0.87 across all 68 markets, every one of them far above the 0.5 that would mark a memoryless coin flip. Calm begets calm and turbulence begets turbulence, across every timeframe we measured. So we expected clustering going in, and the data obliged. Starting from a calm market, the chance of a violent shock in the next three months is about three times lower than starting from an average one, and all 68 markets agree.
The mistake is not in believing the calm. The mistake is in what you do about it.
Here is the sequence, because it matters that it is a sequence. Volatility falls. Your model divides by a smaller number and sizes you up, more than tripling your position. Then the calm persists, because calm begets calm, and while it lasts that large position is working for you. This is not a flaw you feel at the time. It is the reason volatility targeting outperforms in quiet markets. You are being paid, day after day, for carrying size into tranquillity.
And then the state changes. It always does. The volatility that had been falling turns and comes back, and the return that announces the turn arrives while you are holding the largest position the calm ever talked you into. The leverage that paid you for two years takes it all back in an afternoon.
Put the two effects side by side and you can see why the transition is so violent.
Calm shrinks the eventual shock to roughly a third of what it would otherwise have been. Real protection.
Calm more than triples your position size.
The position grows faster than the shock shrinks. So the shock, when it lands, lands on a book that the quiet itself inflated.
The data shows exactly this. The worst levered days do not arrive out of turbulence. They arrive out of below-average volatility, while the book is carrying leverage around a third above its own median. Quiet going in, oversized going in, and then the turn.
“52 of 68 markets suffered their worst levered loss coming out of a quiet market, not a stressed one.”
The calm is not dangerous because it is calm. It is dangerous because it is where you accumulate the position that the transition will use against you.
The specimen case
In the summer of 2018 the Japanese Government Bond market was the quietest major market on earth. Trailing volatility of 0.9% a year, in the 2nd percentile of its own history, close to a flat line. It had been that quiet for a long time, which is exactly why the trouble was coming.
A standard risk model looked at that tranquillity and authorised 16.4 times leverage. For as long as the calm held, that position was profitable, and nothing in the model complained.
Then, on 1 August, the Bank of Japan adjusted its grip on the market and the calm broke. The bond moved six tenths of one percent in a day. In any other context, at any normal position size, nobody would notice a move that small.
At 16.4 times leverage, nine percent of the capital was gone, and the volatility that had been 0.9% going in was more than twice that within the month. The quiet had not been safety. It had been the setup.
The market barely moved. The leverage did the killing, and the calm had loaded the gun. Every step was correct: the position was sized by the book, the risk was measured by the book, and the book mistook a long silence for a permanent one.
3. Diversification that leaves when you need it
Modern Portfolio Theory rests on a promise. Spread your money across things that do not move together and you will be safer than holding any one of them.
The measure of “moving together” is correlation. It is built from squared quantities, so it inherits everything from Episode One.
What should we see if the model is right? Correlations stable enough that the protection you bought is there when you need it.
What we found: correlation rises in a crisis. Every sector. No exceptions.
Across the full 68 markets, crisis correlation is more than triple its calm level.
Measured as protection, roughly half your diversification disappears in a crisis, which is the only time you were ever going to need it.
The efficient frontier is built from a picture of how markets behaved in ordinary conditions. It delivers a portfolio whose central promise expires on the day it is called upon.
4. A number that changes its mind
Beta is meant to describe how sensitive an asset is to the market as a whole. It sets the cost of capital in most corporate valuations ever performed. It is calculated by dividing one squared quantity by another.
What should we see if beta is a real property of an asset? Stability. A market’s beta should not lurch about from year to year any more than its ticker symbol should.
What we found: silver has an average beta of 2.04, and in individual years it has ranged from 0.20 to 7.34. A factor of thirty-six.
And then the finding that should end the discussion.
“23 of 68 markets had a beta that changed sign from one year to the next.”
A number that reverses direction depending on which twelve months you sampled is not describing the asset.
5. The impossible days that keep happening
Black-Scholes prices the world’s options. It assumes markets move in a bell curve with steady volatility.
What should we see if that is right? Extreme days should be almost unimaginably rare, and the model is explicit about how rare.
What we found: every market in the universe has produced a day the model says cannot occur.
The universe is roughly 10¹⁰ years old. Ten billion.
The model does not say these days are unlikely. It says they cannot happen, by a margin so vast the numbers stop meaning anything. 67 of 68 markets have produced one.
The options market worked this out for itself and stopped believing the model without ever abandoning it. Traders now charge more for crash protection than the model says they should. The pattern has a name, the volatility smile, and it has been the market quietly overruling its own textbook since 1987.
6. Judging a manager on noise
From Episode Two: one three-year window in six reports a losing score for a system that genuinely makes money. More than a quarter fall below the level at which managers get fired.
So roughly one manager review in six, conducted on a perfectly sound strategy, concludes that the strategy destroys value.
The pattern across all six
Read them together and one structure appears.
In every case the model is calibrated on the ordinary and then applied to the extraordinary.
The safety limit is drawn using the last 250 quiet days and asked about the next disaster. Leverage is set from the last 60 quiet days and asked to survive the next shock. Diversification is measured in normal conditions and asked to protect a crisis. Beta is fitted on one year and applied to the next. Black-Scholes is calibrated on the middle of the distribution and asked to price the edge.
Squaring concentrates the risk estimate into whichever handful of extreme days happened to fall inside your sample. The events that matter are, by construction, the ones that did not.
These are not faults in the implementations. They are the assumption, arriving on schedule.
Next, in Episode Four: we take forty years of S&P returns and shuffle them like a deck of cards. Every statistic a risk model can see stays where it was. The worst loss changes by fifty percentage points of capital.
Richard Brennan writes on systematic trading, complex adaptive markets, and the philosophical foundations of trend following at atstradingsolutions.com. His books include The Fractals of Finance, Complex Adaptive Markets, Carved by Impossibility and The Aussie Turtles Trend Following Guide.
Want to explore why structure exists at all?
Carved by Impossibility: What Remains When Everything Else Is Eliminated
The book explores the architecture of constraint, emergence, and reality itself, and what it means for how we understand markets, life, and the universe.
Available now on Amazon in paperback, hardcover, and Kindle.
Want the theoretical foundation for why trend following works?
The Fractals of Finance: Determinism, Adaptation and the Geometry of Markets
The book explores the full architecture of feedback, fat tails, and fractal structure in financial markets, and what it means for how we trade, invest, and understand risk.
Available now on Amazon in paperback, hardcover, and Kindle.
Want a practical field manual for trading trends and capturing outliers?
The Aussie Turtles Trend Following Guide: A Field Manual for Hunting Outliers adapts the timeless principles of the original Turtle traders into a systematic, rules-based approach for modern markets. Co-authored with Adam Havryliv.
Available now on Amazon in paperback, hardcover, and Kindle.