Why Can Prices Fall 40% in Seconds Without Human Intervention?

A flash crash is a cascade triggered not by human panic but by algorithmic panic: automated trading systems that sell rapidly in response to price changes, which trigger other automated systems to sell, which trigger more systems, creating a self-reinforcing loop of selling that happens in seconds and evaporates nearly as fast. Flash crashes are qualitatively different from human-driven panics because they have no information content and no delay for thought. A human panic takes minutes or hours to cascade (traders have to see the decline, recognize the fear, and execute the sale). An algorithm cascade takes milliseconds: a price tick down triggers a sell order, which fills at a worse price, which triggers another sell, which cascades through the entire market in 5 seconds. The 2010 flash crash showed that a single large sell order in obscure E-mini S&P 500 futures could trigger a 9% decline in the overall market in minutes, turning trillions in market value to ash and back within an afternoon. Flash crash panic is unique because it is not real; the cascade is mechanical and the prices are false. But the losses it creates are very real.

Quick definition: A flash crash is an extreme, automated sell-off driven by algorithms reacting to price movements rather than information, typically lasting seconds to minutes and reversing nearly as quickly.

Key takeaways

• Flash crashes are triggered by large automated orders hitting thin liquidity, causing prices to cascade downward until the algorithm hits a stop-loss or the order is canceled.
• The May 6, 2010 flash crash showed that a $4.1 billion sell order could trigger a 9% S&P 500 decline in minutes, with some individual stocks falling 40–60% before recovering.
• Flash crashes are self-limiting: once the triggering order is complete or canceled, the cascade stops and prices recover within seconds, showing the crash was mechanical, not information-driven.
• Modern circuit breakers and trading halts have made worst-case flash crashes less likely, but they have not been eliminated; flash crashes occur regularly in individual stocks.
• Retail traders are uniquely vulnerable to flash crashes because they execute market orders during the cascade and fill at catastrophic prices; the best protection is to never use market orders in volatile markets.

The anatomy of a flash crash

A flash crash has a specific structure. First, a large automated order enters the market: "Sell 500,000 shares of SPY" (or an index futures equivalent). The order is so large that it exhausts the available bids at all current price levels. The algorithm is required to break the order into smaller pieces and execute them sequentially. As the first tranche sells, the price ticks down slightly. That tiny price move triggers other algorithms to reassess: Prices just fell 0.5%, my model says they should fall 2% more, I should sell first before that happens. Those selling algorithms enter their orders, pushing the price down further. That move triggers another round of selling algorithms. Within seconds, the price has fallen 5–10%, having exhausted all bids at reasonable price levels.

Now a new phenomenon: panic buying. Traders watching in real-time see prices falling and believe they have found a bargain. They buy aggressively. Their buying exhausts the bids that had been placed by shorters and other fast algorithms. Prices reverse sharply. Within another 5–10 seconds, prices are back to where they started. The cascade has completed in 30 seconds.

The critical insight is that flash crashes have no information anchor. The $4.1 billion sell order in 2010 was not news; it was a large order from a mutual fund with no new information. The cascade was purely mechanical: algorithm → price move → more algorithms → faster price move → exhaustion → reversal. It was a technical failure, not a market failure.

The May 6, 2010 flash crash

The most famous flash crash occurred on May 6, 2010, and provides the definitive case study. That afternoon, a mutual fund (later identified as a Kansas pension fund) wanted to sell $4.1 billion in S&P 500 exposure. To achieve that sale, the fund's algorithm broke the order into 75,000 pieces, to be executed over 5 hours. The algorithm, however, was set to execute the orders at a rate rather than at a price—meaning it would sell the same percentage of volume every minute, regardless of how much the market was moving.

That minor setting had catastrophic consequences. As the algorithm began selling at 2:30 p.m. Eastern time, the S&P 500 futures (which trade with less depth than cash equities) fell rapidly. The algorithm, seeing the price move, did not pause; it continued selling on schedule because its instruction was to maintain a steady execution rate. As futures fell, other algorithms monitoring futures prices (on the assumption that futures lead equities) began selling equities. Those sales hit the market just as the original algorithm was hitting peak execution. The volume of selling for 10 minutes was extreme: trillions of dollars of orders hitting the market simultaneously.

By 2:45 p.m., the S&P 500 had fallen 9%, the Dow Jones Industrial Average had fallen 6%, and individual stocks had fallen 20–60%. Accenture fell from $40 to $1. Sotheby's fell from $30 to $1. Procter & Gamble fell from $60 to $39. These were not small declines; these were cascades that implied the companies had lost 95% of their value in minutes. The moment traders realized something was wrong, they began buying aggressively. Within 5 minutes, prices began to recover. Within 15 minutes, most stocks had recovered to their pre-crash levels. The Accenture shares that filled at $1 were back at $40. The entire cascade lasted 25 minutes.

In the aftermath, the SEC investigated. The findings were clear: the cascade had been triggered by a single large automated sell order, amplified by algorithms reacting to that order, and terminated only when traders realized prices were irrational and began buying. No news had justified the crash. No information had been released. The market had simply broken due to algorithm interaction.

Why algorithms amplify cascades

Algorithms amplify cascades because they have no judgment. A human trader, seeing the market down 5%, might think: This is an opportunity. I should buy. An algorithm, seeing the market down 5%, thinks: My model says prices should move up 2% on low volume, they just moved down 5%, that is a 7% miss, my model is broken, I should not trade until I understand why. That algorithm stops trading. Other algorithms, seeing volume increase and prices fall, think: There must be bad news, I should sell before it gets worse. Those algorithms sell. The cascade accelerates.

Algorithms also exhibit herding behavior, but in a much more synchronized way than humans. If 1,000 traders all have similar algorithms (mean-reversion algorithms that buy when prices are down, momentum algorithms that sell when prices are down), they will all trigger at the same time when prices hit the same level. Humans have cognitive delays and disagree with each other, so their panic is staggered. Algorithms agree with each other perfectly, so their panic is synchronized.

Additionally, algorithms are often programmed to exit trades when losses hit a certain threshold. If 100 algorithms all have stop-losses at a certain price level, they will all execute simultaneously the moment that price is hit. That synchronized selling exhausts liquidity and drives prices lower, hitting the next layer of stop-losses.

Flash crashes versus normal crashes

A flash crash is distinguished from a normal crash by three characteristics:

1. Speed: A normal crash unfolds over hours or days. A flash crash unfolds over seconds to minutes.
2. Reversal: A flash crash reverses most of its losses within 5–30 minutes. A normal crash takes days or weeks to recover.
3. Information anchor: A normal crash is tied to real news (earnings miss, bankruptcy, central bank decision). A flash crash has no news.

Not all rapid declines are flash crashes. During the 2020 COVID crash, the S&P 500 fell 12% in one day (from open to halt), then fell another 20% over the next 10 days. That was a normal crash that happened to unfold over multiple days. During the 1987 Black Monday crash, the market fell 22% in one day, but there was no intra-day reversal; the decline stuck. That was a normal crash, not a flash crash.

A flash crash is specifically the rapid, self-reverting phenomenon triggered by algorithms rather than information. The May 2010 crash is the textbook example.

The role of high-frequency trading in flash crashes

High-frequency trading (HFT) firms operate algorithms that trade thousands of times per second, profiting from tiny price discrepancies. In normal market conditions, HFT provides liquidity: they buy when others sell and sell when others buy, making the spread tighter. However, in a flash crash, HFT algorithms are among the first to detect the cascade and the first to exit. When HFT firms pull their bids (stop quoting prices), liquidity evaporates. The next layer of traders finds themselves in a market with no bids, so they sell at any price. That creates even more selling pressure.

The 2010 flash crash was triggered partly by HFT algorithms reacting to the large mutual fund's sell order. As the sell order hit the market, HFT algorithms detected the pressure, reassessed their models, and began selling along with the fund. This amplified the cascade. When the cascade ended and prices stabilized, HFT algorithms returned and began buying, stabilizing prices and eventually triggering the recovery.

This dynamic has improved somewhat since 2010. The SEC implemented new rules requiring HFT firms to maintain reasonable quotes even during market stress (rule 10b-5). However, HFT remains a significant amplifier of cascades, and flash crashes still occur regularly in individual stocks.

Real flash crash examples

May 6, 2010: The original flash crash: As described, a $4.1 billion sell order cascaded into a 9% S&P 500 decline and 40–60% declines in individual stocks, all within 25 minutes.

August 5, 2010: The dollar flash crash: The U.S. dollar fell 1% against other currencies in seconds, recovering just as quickly. The cascade was triggered by a large automated sell order in a currency pair with thin liquidity. The dollar trades in a 24/7 market with less circuit-breaker protection, so the cascade lasted longer (several minutes) and was more damaging.

March 18, 2020: Treasury flash crash: U.S. Treasury bond yields spiked dramatically on March 18, 2020, during the COVID panic. Bond prices fell 1–2% in minutes as algorithms detected stress and began selling. The cascade was reversed by Federal Reserve intervention (unlimited QE and repo lending) rather than by natural reversal. This shows that flash crashes in fixed-income markets can be more damaging because the assets are used as collateral, and forced selling creates systemic risk.

August 24, 2015: China-triggered volatility cascade: On August 24, 2015, the S&P 500 fell 3.6% at the open, triggering circuit breakers. The cascade was driven by Chinese stock market weakness and automated selling based on currency moves. Several individual stocks fell 10–20% in the first 30 minutes before circuit breakers halted trading.

Protecting yourself from flash crashes

Retail traders are particularly vulnerable to flash crashes because they often place market orders (sell at whatever price is available) during volatile periods. Market orders execute instantly at the worst available bid. During a flash crash, that bid might be 50% lower than fair value.

Protective strategies:

1. Use limit orders, never market orders: A limit order says "sell at $X or do not sell at all." If a flash crash sends prices to $X–20%, your limit order will not execute, protecting you from the catastrophic price. The tradeoff is that you might not sell at all if the cascade is rapid and reverses before hitting your limit. But this is better than selling at the flash-crash price.

2. Do not trade during times of stress: The May 2010 crash happened at 2:30 p.m. Eastern time, during the busiest trading hour. Had you not been trading during that hour, you would not have been affected. Flash crashes are most likely during times of high volatility and stress.

3. Use stops at reasonable levels: If you are going to use stop-loss orders, place them at prices that make sense for your thesis, not at round numbers where other traders' stops cluster. Avoid stopping at the level that triggered everyone else's stop.

4. Understand your broker's protections: Some brokers have "stub quotes" restrictions that prevent extreme prices from executing. If a stock normally trades at $40 and someone tries to sell at $1, the broker may reject the order as a "stub quote." Know your broker's protections.

5. Keep cash reserves: If you hold cash and a flash crash creates irrational prices, you can buy the crash and profit from the recovery. Traders who are 100% invested in stocks cannot exploit flash crashes.

6. Avoid options during stress: Options are particularly vulnerable to flash crashes because their prices move faster than stocks. A stock falling 5% can cause an option to fall 50%. Avoid trading options during market stress.

The post-2010 regulatory response

After the 2010 flash crash, the SEC implemented new rules:

• Circuit breakers on individual stocks: If a stock moves more than 10% in 5 minutes, trading halts for 5 minutes.
• HFT quote requirements: High-frequency traders must maintain reasonable bids and asks even during stress.
• Clearly erroneous trade rules: Trades executed at prices far from the current quote can be canceled (e.g., a $0.01 fill in a stock that traded at $40 is clearly erroneous).
• Order-cancellation ratio limits: Traders who place and cancel orders frequently must maintain a minimum ratio of executed to canceled orders.
• Enhanced position limits: The SEC imposed position limits on certain commodities and index futures to prevent single traders from moving entire markets.

These rules have made worst-case flash crashes less likely, but they have not been eliminated. Flash crashes still occur regularly in individual stocks and less-liquid instruments.

FAQ

How can a market recover from a 40% decline in seconds?

A flash crash recovers quickly because it is not information-driven. The algorithms that sold (sensing a cascade) suddenly detect the reversal and begin buying aggressively. Traders who recognize the crash as mechanical rush in to buy at irrational prices. Within 5–30 minutes, the buying pressure exceeds selling pressure, and prices recover. A normal crash (information-driven) does not recover quickly because traders are reassessing fundamentals, which takes longer.

Can flash crashes cause permanent losses?

Yes. Traders who execute market orders during a flash crash can suffer permanent losses because they sell at irrational prices and do not get the benefit of the recovery (they no longer own the stock). Additionally, if forced selling (margin calls, forced liquidation) coincides with a flash crash, traders may be forced to realize losses at the worst possible moment.

Why do flash crashes happen in some stocks and not others?

Flash crashes are most likely in stocks and instruments with thin liquidity. A large sell order in Apple (very liquid) might move the price 0.5%. The same order in a small-cap stock (thin liquidity) might move the price 10%. The less liquid the instrument, the more vulnerable it is to flash crashes.

Can flash crashes be predicted?

No. Flash crashes are triggered by specific order flow patterns that are nearly impossible to predict. However, you can identify conditions that increase flash crash risk: high volatility, unusual order flow, extremely thin liquidity, and times of market stress.

Do cryptocurrency markets have flash crashes?

Yes, frequently. Cryptocurrency markets have less regulation and thinner liquidity than stock markets, so flash crashes are more common. Bitcoin and Ethereum routinely see 5–10% cascades in seconds. The cryptocurrency market does not have circuit breakers, so the cascades can be severe.

If a flash crash triggers my stop-loss, should I sue my broker?

Maybe. If the broker provided inadequate execution protections and you filled at a clearly irrational price (e.g., a $0.01 fill when the stock was $40), you may have grounds for a complaint with FINRA. However, most brokers now have rules against clearly erroneous trades, and they may cancel your order. The tradeoff is that you get your loss canceled, but you also do not get the recovery.

Related concepts

• How Panic Creates Selling Cascades
• How Circuit Breakers Stop Panic
• The COVID Panic of 2020
• Loss Aversion and the Pain of Selling
• FOMO and Meme Stocks

Summary

A flash crash is an algorithmic cascade triggered by large automated orders hitting thin liquidity, causing prices to fall 5–40% in seconds before recovering just as rapidly. The May 6, 2010 flash crash demonstrated that a single $4.1 billion sell order could trigger a 9% market decline due to algorithm interactions and stop-loss cascades. Flash crashes are self-limiting and information-free: they reverse quickly because they have no relationship to fundamentals. Protecting yourself requires using limit orders instead of market orders, avoiding trading during times of extreme stress, and understanding your broker's protections against clearly erroneous trades. Post-2010 circuit breakers and HFT regulations have reduced the severity of flash crashes but have not eliminated them.

Next

→ The COVID Panic of 2020