How Did a Misplaced Decimal Nearly Destroy the Stock Market?

May 6, 2010. 2:32 PM.

The Dow Jones Industrial Average was down about 300 points. Bad, but manageable.

By 2:47 PM—fifteen minutes later—it was down 998 points.

Nearly a trillion dollars in market value vanished in fifteen minutes.

Then, just as suddenly, it came back.

Investigators spent months trying to figure out what happened.

The answer? A single trade. A bad algorithm. And a cascade of failures that revealed just how fragile our financial system really is.

What Happened: The Timeline

2:32 PM: Markets are jittery. Greece’s debt crisis has everyone nervous.

2:32 PM: A mutual fund (Waddell & Reed) decides to sell 75,000 E-Mini S&P 500 futures contracts. Worth about $4.1 billion.

They use an algorithm to execute the trade. The algorithm has one instruction:

“Sell as fast as possible, but don’t move the market.”

The algorithm interprets this as: “Sell 9% of recent trading volume every minute.”

Seems reasonable.

2:41 PM: High-frequency trading (HFT) firms start buying the contracts. They’re not investing. They’re just flipping them to other buyers for tiny profits (measured in fractions of a cent).

This creates artificial trading volume.

Here’s the problem: the Waddell & Reed algorithm sees this high volume and thinks “market is liquid, I can sell faster.”

So it accelerates.

2:42 PM: HFT firms realize something’s wrong. They stop buying.

Now Waddell & Reed’s algorithm is still selling, but nobody’s buying.

Prices drop.

2:44 PM: Other algorithms notice the price drop. They interpret it as a signal: “Market is crashing, SELL EVERYTHING.”

Cascading sell-off begins.

2:45:28 PM: The E-Mini S&P futures drop 3% in four minutes. That’s insane volatility.

Stock prices on the actual stock exchanges (NYSE, NASDAQ) start crashing too.

Individual stocks hit zero. Apple, Procter & Gamble, Accenture—blue-chip companies—trading for one cent.

2:45:28 PM: Circuit breakers trigger. Trading halts for five seconds.

Five seconds.

That’s all it took for humans to notice and say “wait, this is wrong.”

2:45:33 PM: Trading resumes. Prices snap back. The crash reverses almost as fast as it happened.

3:00 PM: Markets mostly recovered. Dow ends the day down 348 points (bad, but not catastrophic).

Total duration of the crash: 15 minutes.

The Technical Explanation: Liquidity Evaporation

The Flash Crash wasn’t caused by a misplaced decimal (that’s a myth). But it was caused by something just as mundane: bad algorithm design.

Here’s what went wrong:

The Algorithm Didn’t Understand Liquidity

Waddell & Reed’s algorithm was told to execute a $4.1 billion sell order without moving the market.

How do you do that?

You sell slowly. You match your selling speed to available buying demand.

Their algorithm did this, but badly.

It measured liquidity by looking at recent trading volume. High volume = lots of buyers = I can sell faster.

The problem? Volume doesn’t equal liquidity.

High-frequency traders were creating fake volume. They’d buy a contract and immediately resell it. This looked like liquidity, but it wasn’t.

Real liquidity is: “someone who wants to hold this asset.”

Fake liquidity is: “someone who wants to flip this asset in 0.02 seconds.”

When the flippers left, liquidity evaporated instantly.

Feedback Loops

The algorithm selling → HFT firms buying → higher volume → algorithm sells faster → HFT firms leave → prices drop → other algorithms sell → prices drop more → more algorithms sell…

This is a positive feedback loop. Small changes amplify.

In stable systems, feedback loops are negative (self-correcting). Prices drop → bargain hunters buy → prices recover.

But in the Flash Crash, all the bargain hunters were algorithms. And algorithms had been programmed to avoid volatility, not embrace it.

So when volatility spiked, algorithms pulled out. Making volatility worse.

The “Hot Potato” Effect

High-frequency traders don’t hold positions. They’re market makers—they profit from the spread between buying and selling.

During the crash, HFT firms were passing contracts back and forth like a hot potato.

One study found that a single contract was traded 27 times in 14 seconds before finding a buyer who actually wanted to hold it.

That’s not liquidity. That’s chaos.

And the Waddell & Reed algorithm couldn’t tell the difference.

The Human Element: Circuit Breakers Saved Us

The crash stopped because of a five-second circuit breaker.

Five seconds of forced pause.

That’s how fragile the system is. All it took was five seconds for human traders to look at prices and think “Apple is not worth one cent.”

Manual intervention corrected the error.

But here’s the terrifying part: the circuit breaker almost didn’t trigger.

It was implemented in 2008 after previous flash crashes. But the thresholds were set conservatively. The crash had to be severe enough to cross the threshold.

If prices had dropped 9% instead of 10%, the circuit breaker wouldn’t have triggered.

And nobody knows what would’ve happened then.

The Aftermath: We Learned Nothing

After the Flash Crash, regulators implemented reforms:

More circuit breakers (now individual stocks can halt trading if they move too fast)
Limits on how fast algorithms can trade
Requirements for HFT firms to provide liquidity during volatility

Sounds good, right?

Except flash crashes keep happening.

August 24, 2015: Mini flash crash. Dow drops 1,000 points at open, recovers within minutes.

October 15, 2014: Treasury bond flash crash. 10-year yield swings 30 basis points in 12 minutes.

January 2, 2019: Apple flash crashes 10% in after-hours trading due to erroneous sell order.

The market is still fragile. Still dependent on algorithms. Still vulnerable to cascading failures.

The Philosophical Problem: Markets Are No Longer Human

Once upon a time, stock markets were humans trading with humans.

Prices reflected human judgment. Slow. Deliberate. Rational (ish).

Now, 70% of US stock trading volume comes from algorithms.

And algorithms don’t “think” about value. They react to signals.

Price drops → sell.

Volume increases → trade more.

Volatility spikes → pull out.

These rules work 99.9% of the time. But during the 0.1% of the time when markets are stressed, these rules amplify problems instead of solving them.

The Flash Crash revealed something uncomfortable:

Nobody is in control.

Not regulators. Not traders. Not even the people who wrote the algorithms.

The market is a complex system. And complex systems have emergent behaviors—properties that arise from interactions, not from any single component.

The Flash Crash was emergent. No one intended it. No one predicted it. And no one knows how to prevent the next one.

The Math: Why This Keeps Happening

Markets are supposed to be self-stabilizing.

Price drops → bargain hunters buy → prices recover.

That’s negative feedback. Economics 101.

But algorithms introduce positive feedback.

Price drops → algorithms sell → prices drop more → more algorithms sell.

And once you have positive feedback, you get instability.

The math here is simple:

dP/dt = -k(P – P₀) + A(P)dP/dt

Where:

P = current price
P₀ = equilibrium price
k = rate of human correction (negative feedback)
A(P) = algorithmic response function (positive feedback)

When A(P) is small, the system stabilizes.

When A(P) is large (lots of algorithms trading), the system becomes unstable.

And the instability grows exponentially with the number of algorithms.

That’s why flash crashes have gotten more frequent as algorithmic trading has increased.

Not because algorithms are “worse.” But because more algorithms = stronger positive feedback = more instability.