Summer series: Tales from Jobs Past
“We were in a war room contemplating if we should call lawyers,” recalled Timothy James, a product technology manager at Google’s Bakery Square campus in Larimer. “For five hours, we were convinced we did something wrong and were going to jail.”
Soon after Mr. James began working in algorithmic trading at Thomson Reuters, where he would work until joining Google in 2013, he experienced the Flash Crash of 2010 firsthand — a $3 trillion stock market crash that lasted only 36 minutes but led to pandemonium in his office and throughout Wall Street.
“The market dropped rapidly,” Mr. James said. “Basically, if you think about it, there’s a bunch of people saying they want to buy at a certain price and a bunch saying they want to sell at a certain point. But if everyone willing to buy is wiped out, the market moves drastically.”
And that’s exactly what happened. Prices for some stocks fell to one penny.
As an algorithmic trader who programs computers to follow a set of rules for placing automated stock market trades, Mr. James knew how quickly errors could multiply.
Read More
“If you’re looking at the market and trying to determine whether to buy or sell and you see a bunch of people selling ... people think there’s a reason to sell, so they do, too,” he said.
Computers work the same way through algorithms that teach them when to buy and sell. Those transactions occur at a speed and frequency that human traders cannot compete with, meaning Mr. James and the rest of his team were spooked when they thought a bug in their system accidentally caused the crash.
May 6, 2010, began as a normal day, Mr. James said, and he remembered it was a Thursday because he was set to have a relaxing Friday the next day — working remotely from home, rather than reporting to the office in Sewickley or traveling to the headquarters in New York City. Instead, Friday included a long conference call with bated breath, trying to determine how to react to the crash.
Not long before the stock market crash on that Thursday, Mr. James’ team had a client testing one of its applications. The client accidentally dropped a book on the keyboard “and it began sending order after order after order,” he said.
While those products were still in beta testing, the algorithm trading team wondered if the client had accidentally moved them to the live market with bugs intact.
“The early rumor that day was that one of our customers had placed a large trade for 6 billion instead of 6 million,” Mr. James said. “It was several orders of magnitude off ... we knew there was a bug in our system that customers had encountered before.”
He explained this is called a “fat finger trade,” or a keyboard error in financial markets. In practice, if a customer wants to place an order for $1 million, he or she types “1M.”
If someone accidentally typed in “1B,” however, the fat finger order would consume far more liquidity than intended, placing an order at $1 billion.
“We were trying to do damage control ... we were trying to see if this was something our software could even cause,” he said. “We were trying to replicate the problem ... within a few days of testing, we saw it wasn’t something caused by our algorithms.”
Still, it took days to digest what had actually occurred that day, and it took years for the government to find the culprit.
In 2015, Navinder Singh Sarao, a trader living in London, was charged with using spoofing algorithms — a disruptive form of trading wherein an algorithm bids or offers with the intent to cancel before orders are filled. In 2016, he pleaded guilty.
Before the Flash Crash, he had placed thousands of E-mini S&P 500 stock index futures contracts and planned to cancel them.
If a stockholder wants to buy more of a given share when it costs $12, but wants to sell when it hits $15 per share, “Then there’s a lot of space in the middle,” Mr. James explained. “If you’re a large fund, you’re looking at a lot of different instruments. You might run an algorithm that says you buy at $14 ... as soon as that happens you make the transaction.
“It shows everyone this person is willing to buy at that price [of $14]. Then you might send in another order at a slightly lower price and keep doing that until you find the best price,” he said.
“The problem is ... if someone is willing to sell at $13, they might not want that order to be public because they’d prefer to sell at $14 or $15.”
Through that process, a spoofer can artificially drive down prices, leading the market to plummet. Then, the spoofer can buy a bunch of shares, eliminate the algorithm and profit by selling the shares bought at a lowered price a few minutes before, once trading returns to normal.
Needless to say, spoofing is now illegal, and Thomson Reuters was not the cause of the crash.
“It was surreal being in a position where we thought we had caused a global financial crash,” Mr. James said, now that he can laugh about it. “It was not the relaxing Friday working from home that I thought it would be.”
Courtney Linder: clinder@post-gazette.com or 412-263-1707. Twitter: @LinderPG.





Comments
Welcome to Post-Gazette subscriber-only commenting community. Please review our community guidelines.