2010 Flash Crash triggered by automated sell algorithm
An automated sell program dumped $4.1 billion of E-mini S&P 500 futures in about 20 minutes without regard to price or time, contributing to a roughly 998-point intraday drop in the Dow.
What happened
On May 6, 2010, at approximately 2:32 p.m. EDT, a large fundamental trader (Waddell & Reed Financial) initiated an automated sell program to sell 75,000 E-mini S&P 500 futures contracts, valued at approximately $4.1 billion. According to the CFTC-SEC joint report "Findings Regarding the Market Events of May 6, 2010" (issued September 30, 2010), the program was set "to target an execution rate set to 9% of the trading volume calculated over the previous minute, but without regard to price or time." A sell order of that size would normally take about five hours to execute, but the algorithm completed the sale in roughly 20 minutes. The Dow Jones Industrial Average fell about 998.5 points (roughly 9%), with most of the decline occurring within minutes before largely recovering; the overall event lasted about 36 minutes. Press coverage characterized the episode as temporarily erasing on the order of $1 trillion in market value, a figure that does not appear in the CFTC-SEC report.
What the agent did
An automated execution algorithm placed and completed the $4.1 billion sell program. The algorithm's volume-participation logic, targeting 9% of the prior minute's trading volume while ignoring price and time, drove the pace of selling. The decision to run the program was made by humans at the fundamental trader; the consequential action of rapidly executing the trades into a falling market was carried out by the automated system.
The irreversible effect
Roughly $4.1 billion in E-mini S&P 500 futures were sold into the market in about 20 minutes, contributing to a rapid roughly 9% intraday decline in the Dow Jones Industrial Average and severe short-lived dislocations across equity and futures markets. While prices largely recovered within the hour, the executed trades and the resulting market disruption could not be undone.
Root cause
The execution algorithm was configured to participate at 9% of recent trading volume without any price or time limits. In thin, already stressed conditions its own selling increased volume, which caused the algorithm to sell faster in a feedback loop, executing the entire large order far faster than intended and driving prices sharply lower.
How a maker-checker control would have refused it
A maker-checker control would not have stopped this in the moment, because a human maker chose to run the program and the harm came from an unattended algorithm executing at machine speed. The relevant control is a hypothetical limit or gate: the sell program had no price limit, no time constraint, and no participation cap tied to available liquidity, so it ran without regard to price or time. A pre-trade limit check requiring an explicit price or rate ceiling, or a high-risk gate that flags an order representing an outsized share of expected volume for a $4.1 billion notional, could have capped the execution pace or paused it for human confirmation before it cascaded.
Runnable reproduction
A runnable reproduction for this entry is in progress.
Accuracy and corrections
This entry describes a publicly reported incident and is compiled from the primary sources listed above. Where an account is a legal allegation rather than an established finding, the entry labels it as such. Summaries can still contain errors. If you can document a correction, email hello@makerchecker.ai and we will review and correct it, with the change noted, within 14 days.
See it for yourself
Reading is one thing. Watch it block an agent.
One command starts the demo: an agent stopped from signing off its own work, and the signed evidence file an inspector can check for themselves.
Designed against the rules your auditors already enforce.