When Genius Failed Ch. 5: They Were Right — The Trades Converged After the Fund Died

Investment Background

The most widely repeated claim about LTCM is "their models were wrong."

In the sense that matters most, that claim is false — and correcting it is the most valuable thing this book can give you.

Here is the fact:

The fourteen-bank consortium that took over LTCM unwound those positions over the following year and more. They made money doing it.

By the time the fund completed its wind-down in early 2000, the rescue consortium had recovered its capital with a positive return. The positions supposedly proven "wrong" had, in the overwhelming majority, converged exactly as the original logic said they would.

On-the-run/off-the-run spreads normalized. Italian-German government bond spreads narrowed. Swap spreads returned to their historical range.

LTCM's judgment about the market was, in hindsight, substantially correct.

It still lost investors 92% of their money.

The Wall Street Translation

This fact deserves very serious attention, because it overturns a belief most investors live by:

"If my analysis is right, I will make money eventually."

That sentence is false.

The correct version is:

"If my analysis is right, AND I can hold until it is proven, I will make money."

The second condition is not a footnote. It is exactly as important as the first, and it is harder to satisfy.

LTCM satisfied the first and failed the second. The outcome was indistinguishable from being completely wrong.

Quantifying It

Let us turn "how long can you hold" into something calculable, because that is what makes it executable.

Suppose you judge an asset to be undervalued and intend to hold until it reverts.

The key question is not "will it revert" but "how far can it fall before it reverts, and can I survive that."

For an unlevered holder:

• Asset falls 50% → 50% paper loss → you still own every share
• Asset reverts → you recover fully and profit

For a 25x levered holder:

• Asset falls 4% → your capital is zero and the position is liquidated
• Asset reverts → irrelevant to you; you are no longer present

The same correct judgment, two opposite endings. The difference lies not in the quality of the judgment but in the capacity to endure.

This is why Way of the Turtle in this library spends an entire chapter on position sizing. Position sizing is not a technique about how much you make. It is a technique about how long you last.

A Brutal Arithmetic About Time

Keynes's endlessly quoted line:

"The market can remain irrational longer than you can remain solvent."

LTCM is the most expensive empirical demonstration of it ever produced.

Note the structure of the sentence. It compares two durations:

1. How long the market takes to correct an error — outside your control.
2. How long you can avoid being forced out — entirely within your control.

Most investors spend all their effort forecasting the first. This is futile.

The professional approach is to maximize the second.

And maximizing the second is remarkably simple, requiring no forecasting ability at all:

• Use no leverage → your available time becomes unlimited
• Hold sufficient cash → you need not sell at lows to fund living expenses
• Avoid instruments that expire → options run out; your shares do not

Anyone can execute those three. None requires you to be smarter than the market.

Connections to the Rest of the Library

This chapter is where several threads across the library converge, and the links are worth stating.

Book	What it says	What this chapter adds
The Intelligent Investor	Mr. Market quotes absurd prices; you may ignore him	Only the unlevered are entitled to ignore him. The levered must accept his quote.
Antifragile	Build a structure that gains from volatility	That structure presupposes you are alive through the volatility
Way of the Turtle	Size positions by volatility	The real constraint on size is not expected return; it is survival probability
Winning the Loser's Game	Time is your one structural edge over institutions	Leverage erases that edge outright

The Ellis line is especially important. He argues that the individual investor's one genuine advantage over institutions is that nobody withdraws your capital after a bad quarter — you can hold for twenty years.

LTCM proved how fragile that advantage is: the moment you borrow, you hand it to your creditors.

So — Is the Lesson "Don't Trust Models"?

No. That is another common misreading, and it is equally harmful.

LTCM's models were accurate at the question they were designed to answer. They estimated spread behavior under normal market conditions accurately.

The problem is that the models were used to answer a question they were not designed for: "how much leverage can we run?"

The answer to that question does not depend on volatility under normal conditions. It depends on volatility under worst-case conditions — and worst cases are, by definition, sparsely sampled in historical data.

Thirteen books in this library discuss overfitting and walk-forward validation. What this chapter adds is: even if your model is not overfitted at all, even if it performs honestly well on historical data, it still cannot tell you what happens in a situation that has never occurred.

The correct posture is:

Use the model to estimate normal conditions. Use "what if the model fails entirely" to set position size.

Executable Trading Rules

1. For every position, evaluate two things separately: the probability your judgment is right, and how long you can hold. Most people do only the first. The second is what determines whether you ever collect on the first.

2. Calculate your holding horizon and write it down. Concretely: assume this position falls 50% from today and stays there for three years. Ask — will I be forcibly sold? Will I need to sell for cash? Will I sell under psychological pressure? Any "yes" means your size is too large.

3. Treat cash as a purchase of time. The cost of holding cash is forgone return. What it buys is the right not to sell during an adverse period. For a retiree that right is extremely valuable, because withdrawals are continuous.

4. Never hold a correct judgment in a form that expires. If you believe an asset is undervalued, buy the asset — not a call option on it. Options expire; undervaluation can persist longer than the option. This is the most direct individual-level application of the LTCM lesson.

5. Accept that "I was right and I lost money" is a real and common outcome. It is not bad luck; it is structure. If it happens to you, examine your size and your horizon — not your analysis.

Relevance to a Retirement Portfolio

For a retirement portfolio the implication is concrete, and it maps directly onto the withdrawal tools on this site.

A retiree faces the identical structural problem LTCM faced — only in a different form.

A retiree must withdraw living expenses every year. That means when markets fall, they are compelled to sell assets. This is a mild version of forced selling, but structurally the same thing.

That is precisely the nature of sequence-of-returns risk, which we work through numerically in Chapter 2 of Retirement Decumulation Mechanics.

And the solution is the same one: ensure you do not have to sell at the worst time.

Concretely, hold one to three years of essential spending in cash or short-term bonds as a buffer. That buffer does not raise returns — it slightly lowers long-run returns.

What it does is buy time.

LTCM proved with $4.6 billion that right without time equals wrong. And time can be bought with cash.

Chapter 6 sets out where this framework stops, and the most common mistaken conclusions drawn from LTCM.