LTCM: How Two Nobel Laureates Nearly Broke the System

Two Nobel Prize winners. A former Vice Chairman of the Federal Reserve. The most feared bond traders Wall Street had ever produced.

They built a fund holding roughly $125 billion in assets and over $1.25 trillion in derivatives.

In the autumn of 1998, it lost about $4.6 billion in five weeks — and the Federal Reserve Bank of New York put the heads of Wall Street’s largest firms in a room and told them to fix it before markets opened.

Here is the part that should bother you.

The risk formula that destroyed them is still running. In every bank. In every clearinghouse. In the margin system of the app on your phone.

The team that shouldn’t have existed

1993. John Meriwether — the legendary bond trader from Salomon Brothers — assembles a group that reads like a screenwriter’s idea of finance.

Myron Scholes and Robert Merton, who in October 1997 would win the Nobel Prize in Economics for the options-pricing formula that supposedly made risk calculable. David Mullins, former Vice Chairman of the Federal Reserve. And the sharpest arbitrage desk Salomon ever assembled.

They called it Long-Term Capital Management.

Wall Street CEOs invested their personal money. Banks queued up to lend on terms they extended to nobody else — in many cases charging zero initial margin, a privilege normally reserved for sovereign governments.

The returns justified it. 43% in 1995. 41% in 1996.

One year after the Nobel, the formula destroyed them.

The strategy, in two cans of Coca-Cola

Imagine two identical cans of Coke. Same product, same company, same taste.

One costs $1.00 at the corner shop. The other costs $1.02 at the petrol station next door.

You buy the cheap one. You sell the dear one. You pocket two cents.

That is convergence arbitrage, and it is what LTCM did — with government bonds rather than soft drinks. Buy the marginally cheaper bond. Short the marginally dearer one. Wait for the gap to close, as it almost always does.

The logic is sound. It is, in fact, close to riskless — in the sense that the two bonds really are nearly identical.

But two cents is nothing.

So to make real money, you borrow. Enormously.

LTCM ran, at times, more than 25:1 on its balance sheet — roughly $125 billion of assets — plus over $1.25 trillion of off-balance-sheet derivatives.

They bought $125 billion of Coca-Cola. On credit.

The spark, and then the fire

17 August 1998. Russia defaults.

LTCM’s direct exposure to Russia was survivable. That was never the point.

Russia was the spark. What followed was the fire.

Global panic. Every investor on the planet running for the same exit at the same moment — dumping anything that smelled of risk, buying anything that smelled of safety.

In the Coca-Cola analogy: suddenly nobody will touch the corner-shop can. Everyone wants the petrol-station can, and only that one.

The price gap that was supposed to narrow — exploded.

Their positions were designed to profit from convergence. The market delivered divergence, everywhere, at once.

The number the model said was impossible

The President’s Working Group later recorded the arithmetic: LTCM held $4.1 billion of capital on 31 July 1998, and lost a further $1.8 billion during August alone — taking the equity loss for the year past 50%.

Their internal model had a view on this. A loss of that magnitude in a single month was, by their own calculation, a 10-sigma event.

Let me translate what a 10-sigma event means.

According to the model, it should not occur once in the lifetime of the universe. Not once in 13.8 billion years.

It happened in the fund’s fourth year of operation.

And this is where leverage becomes a death sentence

Watch what happens next, because it is the mechanism that kills every leveraged institution, and almost nobody explains it.

The assets barely moved. Still roughly $125 billion.

The equity underneath collapsed. $4.8bn → $2.3bn → a few hundred million.

Leverage is a ratio. When the denominator evaporates and the numerator does not, the ratio does not drift. It detonates.

25:1. Then past 50:1. Then past 100:1.

By late September, with a few hundred million dollars of capital remaining, LTCM was reportedly running at roughly 250:1.

At 250:1, a 0.4% move against you erases everything you have.

Not a crash. Not a catastrophe. Four tenths of one percent.

The room, and who refused to pay

23 September 1998. The Federal Reserve Bank of New York convenes the heads of Wall Street’s largest institutions.

The following day, fourteen firms put in $3.6 billion.

Bear Stearns — LTCM’s own clearing broker — refused to contribute a cent. (Ten years later, Bear Stearns would need rescuing itself. Make of that what you will.)

And here is the detail that gets misreported constantly:

Not one dollar of taxpayer money was used.

The banks paid. With their own capital. Not out of altruism — out of self-interest, because a disorderly liquidation of $125 billion of positions into a panicking market would have destroyed their balance sheets too.

The Fed organised the room and applied the pressure. That is all it did. Alan Greenspan defended exactly that framing before Congress a week later.

Now the formula — because it is still in your life

Forget the fund. This is the part that matters.

Value at Risk is a weather app

Imagine a weather app that only looks at the last fortnight. Fourteen sunny days, so it tells you: 95% chance of sunshine tomorrow.

It does not tell you what happens in the other 5%. It does not mention that the 5% might be a hurricane.

And here is the genuinely dangerous part: if everybody uses the same app, and the app says sunny, then everybody leaves the umbrella at home. When the storm arrives, everyone gets soaked simultaneously.

That app is called Value at Risk. VaR. Nearly every major bank, broker and clearinghouse runs a version of it. It takes one or two years of market data, runs it through a statistical model, and produces a single number: the most you should lose on a normal day.

It breaks in three places, and the three compound.

Break one: markets are not a bell curve

VaR assumes a normal distribution. Small moves are common; extreme moves are nearly impossible.

Markets do not read statistics textbooks. Extreme moves happen far more often than the bell curve permits — which is why a fund run by the people who invented modern options pricing was able to experience an event their own model priced at once-in-the-lifetime-of-the-universe.

The maths was not wrong. The assumption underneath the maths was wrong.

Break two: it makes crises worse — procyclicality

This is the one that should genuinely alarm you.

In calm markets, volatility is low, so VaR reports low risk — so institutions borrow more, and build bigger positions, because the model says there is room.

Then a shock lands. Volatility explodes overnight.

Now the same model that said all clear yesterday screams SELL — to every institution, at the same time, because they all use the same kind of model.

They all sell into the same market at the same moment. Prices fall further. Volatility rises again. VaR rises again. The model screams louder.

That is the doom loop. The risk model does not prevent the crash. It manufactures it.

Break three: it assumes someone will buy

VaR assumes you can liquidate at today’s price.

When you hold $125 billion in a panicking market, there is no bid. The model pretends there is.

Two things people get wrong about LTCM

“It was a black swan. Nobody could have seen it coming.”

No. When you run 250:1 leverage on correlated positions and never stress-test for a liquidity shock, a wipeout is not an anomaly — it is arithmetic waiting for a date.

“They had Nobel Prize winners. The models must have been sound.”

The models were computed correctly. They were built on the assumption that markets behave like physics.

Markets are not physics. Markets are people running for the same exit at the same time.

There is one further detail from the official post-mortem that ought to be taught in every risk course. LTCM had 75 lenders and over 50 derivatives counterparties. Not one of them knew the fund’s total exposure. There were no covenants limiting aggregate leverage, and no mechanism anywhere to add it all up.

They thought they were lending to gods. So they switched off their own alarm system.

What this means for your portfolio

CHECK — if you use margin, you are inside this machine

Open your broker app. Find maintenance margin or portfolio margin.

Those numbers are produced by VaR-type models. In a volatility spike they rise automatically — and if you fall below the threshold, your positions can be liquidated at the worst possible moment, at the worst possible price, without anyone calling you.

The same logic sits inside risk-managed robo-advisers: strategies that cut exposure when a volatility threshold is breached are, by construction, selling into falling markets. That is not a scandal — it is what they promised to do. But you should know it is what they do.

AVOID — two things

Daily-reset leveraged ETFs, unless you can explain volatility drag to another person without looking it up. They rebalance every day, which makes them procyclical by design and can grind you down in a choppy market even when your directional call was right.

And avoid believing your diversified portfolio is diversified. LTCM was spread across eight countries. On paper, textbook global diversification. In reality every position was the same bet: short liquidity, long credit risk. When the panic came, correlations went to one and the diversification evaporated exactly when it was needed.

Diversification across countries is not diversification across risk factors.

BUILD — a liquidity buffer, and be honest about why

Cash, or short-dated government bonds. Unleveraged. Boring. Slightly embarrassing to hold in a bull market.

It is not dead money. It is a call option on the next crisis.

When the leveraged investor gets a margin call and is forced to sell at a fire-sale price, someone is on the other side of that trade. The only question is whether it is you.

And if you think this is history

Right now, hedge funds are running the same convergence trade — cash Treasury bonds against Treasury futures — funded overnight in the repo market, where the funding can vanish in a day.

In a FEDS Note published on 22 June 2026, Federal Reserve staff sized the basis trade at roughly $830 billion of hedge funds’ Treasury long positioning — about double its previous peak in early 2020 — with hedge fund net repo borrowing near $1.8 trillion at the end of 2025.

Same structure. Same funding fragility. In the market that prices every other asset on earth.

Two Nobel laureates. The smartest traders alive. A formula that said the disaster was impossible.

The formula was right — mathematically.

But markets are not mathematics. They are human beings, running for the same exit, at the same time.

And the formula that missed it is still running. In every bank. In every risk department. In the margin system of your broker.