The 2010 Flash Crash
On May 6, 2010, US stocks fell nearly 9% and recovered within about half an hour — a single large sell algorithm, thin liquidity, and a cascade of automated reactions turned an ordinary bad afternoon into the fastest crash and recovery in market history.
At around 2:32pm on May 6, 2010, a mutual fund began executing a large sell order in E-mini S&P 500 futures using an algorithm that sized itself off recent trading volume, without regard to price or time. As the algorithm sold aggressively into a market that was already nervous about the Greek debt crisis, high-frequency trading firms who normally absorb that kind of selling started passing the contracts to each other rather than holding them, briefly creating a "hot potato" effect that drained the depth of the order book. Liquidity in the futures market evaporated in seconds, and the effect spread almost instantly to the underlying stocks.
By 2:45pm, the Dow Jones Industrial Average had fallen nearly 1,000 points intraday — a drop of about 9% — before recovering almost all of it within roughly twenty minutes. Some individual stocks traded far more violently than the index: shares of well-known companies briefly executed at a penny or as high as $100,000, prices with no relationship to their value moments before, because in some venues the order book had emptied out and stub quotes (placeholder orders far from the market, never meant to actually trade) were the only thing left to match against.
What made it cascade
Three things compounded each other. First, the sell algorithm kept feeding orders based on volume rather than price, so it sold harder as prices fell, rather than pausing. Second, the market-wide circuit breakers that exist today didn't exist yet in their current form — individual exchanges had different rules for when to pause, so an order routed from one venue to another could arrive when its origin had halted but its destination hadn't, executing against a near-empty book. Third, many participants' own risk systems reacted to the chaos by pulling their standing quotes entirely, which is individually rational — nobody wants to be the only one still quoting into a crash — but collectively removed the very liquidity that would have limited the damage.
What changed afterward
Regulators introduced Limit Up-Limit Down bands and updated market-wide circuit breakers, both designed so that a stock's price is walled into a moving corridor and trading pauses briefly rather than executing at absurd prices. Exchanges also adopted clearly-erroneous-trade rules to formally bust trades that occurred at prices far outside the norm during the crash, and single-stock circuit breakers were introduced as an interim fix before LULD was finalized. The event remains the reference case for why US equity market structure needs automatic, price-aware pauses rather than relying on human judgment or thin liquidity to hold up under stress.
The 2010 Flash Crash showed that a large, price-insensitive sell algorithm combined with high-frequency traders passing inventory back and forth rather than absorbing it can drain order book depth in seconds, letting prices detach from fundamental value until trading is forcibly paused — which is exactly what Limit Up-Limit Down and modern circuit breakers were built to do.
It's tempting to blame the crash entirely on "high-frequency trading" as a single villain, but the SEC/CFTC report is clear that no single actor caused it — a conventional institutional sell algorithm started it, and the crash was a systemic interaction between that order, thin liquidity, and everyone else's automated reactions to the chaos, not one rogue algorithm gone haywire.
Related concepts
Practice in interviews
Further reading
- SEC/CFTC, Findings Regarding the Market Events of May 6, 2010
- Kirilenko, Kyle, Samadi & Tuzun, The Flash Crash (2017)