On May 6, the stock market crashed–for about one minute. The Dow Jones Industrial Average fell 998.5 points, one of the biggest declines in history, and then rocketed back for the second largest swing ever at more than 1,000 points. Known as the Flash Crash, the event quickly spawned some witty T-shirts (“I survived the crash of 2:45”), and more importantly, lots of questions over how such a dramatic dip could’ve occurred and so quickly disappeared. Chicago-based data firm Nanex may have solved the mystery, with the help of surprisingly beautiful visuals.
Nanex founder Eric Scott Hunsader believes high-frequency traders may be to blame for the crash. These traders, using advanced equipment and complicated algorithms, issue and cancel thousands of quotes on stocks every second, without ever intending to execute on those prices. This illegal process, referred to as “millisecond quoting” or “quote stuffing,” enables traders to manipulate stock bids and offers, raising them as much as 10% higher, according to Zero Hedge. “These guys believe it’s okay to put in 9,000 quotes and cancel them in one second, for one stock,” Hunsader says. “Nobody is looking at that timeframe, so it just gets lost. It becomes a blip.”
The difficulty then becomes: How does one track these illegal quotes if they occur thousands of times within milliseconds?
About a week before the Flash Crash, Hunsader started noticing patterns in the data. “On April 28, at 11:23 in the morning, you had the same stocks, like Proctor & Gamble, Walmart, all these big, widely held companies with huge capitalization, drop 50 cents in about a second, and then recover,” he says. After pointing out the strange patterns to colleagues, Nanex developer Jeffrey Donovan began plotting the quotes in charts, and soon wrote software to scan for similar data trends.
Soon, Hunsader and Donovan realized they may have uncovered the Flash Crash’s through these visuals.
“A quote really has four components: bid price, bid size, ask price, and ask size,” Hunsader explains. “We found that when there are 5,000 quotes in one second for a stock, there’s a wide-variety of patterns showing those four things changing. So you might get a case where just the bid price changes a certain way. It might go from $20 to $40, a penny at a time, every millisecond. Or there’s patterns that don’t repeat until there’s been 500 quotes. So after 500 of them, the pattern repeats again.”
Plotting this copious data on charts allowed Nanex to see the bigger picture, which might have gone undetected without a visual aid–just another blip lost in another millisecond.
Hunsader and Donovan now refer to these odd patterns as crop circles, and have devoted a page of Nanex’s Web site to the Market Crop Circle of the Day, which includes an RSS feed. They label the pieces like works of art (eg, “Tesla’s Cathedral” (seen below) and “Eraser Head”) and have been able to find dozens if not hundreds of examples of “quote stuffing” any given day.
“When I first suggested to Donovan that he plot the data, I’m not sure he immediately understood,” says Hunsader. “Because a couple hours later, he was just in shock, like, Oh my god!”
“A picture truly is worth a thousand words.” Not to mention a thousand quotes.