Every Edge Has a Painful Period (And That’s Why It Still Works)

Joel Greenblatt published a formula that beat the market, told everybody exactly how it worked, and watched it continue to beat the market.

That should be impossible. An edge announced in a bestseller ought to be arbitraged into nothing within a year.

His explanation is the most useful sentence in value investing, and it applies to every strategy you will ever run. The formula works because it does not work for years at a time. Nobody can hold it through the stretch where it lags, so almost nobody collects.

The underperformance is not a flaw in the edge. It is the toll booth guarding it.

How much of your life is spent underwater

Take a genuinely profitable system. Forty percent win rate, 2R winners, one percent risk, twenty trades a month. Expectancy +0.20R. Over a year it makes serious money.

Now ask a question nobody asks: what fraction of that year is spent below the highest equity you have ever seen?

Eighty-three percent of the time, a winning trader is looking at a number smaller than one he has already seen.

Read the two right-hand columns together, because they are the whole story. Over twelve months this trader almost certainly makes money, finishing below his starting balance only 2.3% of the time. And he spends five days in six feeling like he is losing.

The dry spell, measured

Averages soften this. What matters is the single longest stretch, because that is the one that ends your relationship with the strategy.

Over three years of trading, how long is the worst continuous period below the high-water mark?

Six months. Half a year of watching a positive-expectancy system fail to make a new high, executed perfectly, with nothing wrong.

And one trader in twenty spends more than thirteen months there. Over a year, with a good system, doing everything right.

Nobody plans for this. Every trader plans for the drawdown, quantified in percent. Almost nobody plans for the duration, and duration is what breaks people. A twelve percent drawdown that resolves in three weeks is a story you tell later. The same twelve percent, grinding sideways for eight months, is how a trader arrives at the conclusion that the market has changed.

Why the edge survives being published

Now assemble the pieces.

An edge is arbitraged away when enough capital runs it. Capital runs it when the people holding it can tolerate what holding it feels like. And what holding it feels like is eighty-three percent of your life below your own high, punctuated by a six-month stretch of nothing.

So the edge is not protected by secrecy. Greenblatt gave it away. It is protected by tracking-error tolerance, which is the willingness to look wrong, publicly, for long enough that the arithmetic can finish.

That capacity is scarce. It is scarcer than intelligence, scarcer than information, and it cannot be bought, borrowed, or automated, because the person who abandons the system is the person operating it.

How strategy-hopping manufactures negative expectancy

This is the mechanism, and it is worth stating precisely, because it explains a trader who does everything right and loses.

He runs a +0.20R system. It enters a normal dry spell. Four months in, he concludes it is broken and switches to a different +0.20R system.

The new system, being new, is at its own high-water mark. It promptly enters its own dry spell, because they all do, eighty-three percent of the time.

He has now taken two positive-expectancy systems and constructed, out of them, a process that systematically exits each one during its worst period and enters the next one at random. He collects the drawdowns and never the recoveries.

Strategy-hopping is the mechanism by which a positive-expectancy trader realises negative expectancy. Not bad systems. Good systems, sampled at exactly the wrong points, by a person responding rationally to information that contains nothing.

What to write down before you trade

Your expected underwater period. Simulate your own system: win rate, average R, risk fraction, ten thousand runs. Record the median longest dry spell and the ninety-fifth percentile. Six months and thirteen months, roughly, for the system above. Yours will differ. Knowing it is the difference between enduring a dry spell and discovering one.

The conditions under which you are permitted to change the system. Written in advance, dated, and requiring more than a feeling. A drawdown deeper than your ninety-fifth percentile. A structural change you can name. A rule violation you can point to. Nothing else counts, and “it stopped working” is not a condition, it is a mood with a hypothesis attached.

The distinction between broken and quiet. A broken system produces losses outside its own historical distribution. A quiet system produces nothing outside anything. Most traders cannot tell these apart because they never established the distribution, and so every quiet period is indistinguishable from a break.

The bargain

The edge exists because the experience of holding it is unpleasant. Remove the unpleasantness and you remove the edge, because everybody would hold it, and it would be priced.

Which means the discomfort is not the obstacle standing between you and the returns.

It is the thing you are being paid for.