'Rogue' Algorithm Blamed for Historic Crash of the British Pound

Image: TheBigData
Image: TheBigData

The British pound suffered a “flash crash” earlier this morning in which it plummeted six percent against the US dollar within a matter of minutes. All signs point to high frequency stock trading as the culprit—and possibly a single algorithm.

Advertisement

The British pound fell by more than six percent this morning, suddenly dropping to $1.18, a value that hasn’t been seen in 31 years. The currency quickly rebounded, finally settling in at about $1.24. The so-called “flash crash” was brief, but it’s still sending shockwaves throughout the financial world.

“I initially doubted what I saw on my screen,” said Kenji Yoshii, a foreign exchange strategist at Mizuho Securities, in the Wall Street Journal.

Last night’s dramatic tumble. (Image: Wall Street Journal)
Last night’s dramatic tumble. (Image: Wall Street Journal)

Normally, dramatic drops like this are triggered by major news events, such a declaration of war or a monumental political development. But experts say this incident was likely the result of trading algorithms that were reacting to recent comments made by French President Francois Hollande, who called for tougher Brexit negotiations.

“Apparently it was a rogue algorithm that triggered the sell-off after it picked up comments made by the French President, Francois Hollande, who said if Theresa May and [company] want hard Brexit, they will get hard Brexit,” noted Kathleen Brooks, research director at City Index.

Advertisement

She says that some modern algorithms trade on the back of news sites, and even on what’s trending on social media sites like Facebook and Twitter. “[A] deluge of negative Brexit headlines could have led to an algo taking that as a major sell signal for GBP,” said Brooks. “Once the pound started moving lower than more technical algos could have followed suit.”

Advertisement

High frequency stock trading is a form of rapid-fire trading that involves algorithms, or bots, that can make decisions on the order of milliseconds. They’re guided by factors such as time, price, some fancy math—and even headline news. Compared to these lightning-fast traders, humans are slower by an order of magnitude, which means we’re increasingly being left out of the loop. Stock trading represents the first major domain in which we’re getting AI to do most of the work, and an entirely new digital ecology is emerging.

But as this latest incident shows, mistakes do happen. This “rogue” bot (or bots—we still don’t know what actually caused this latest flash crash) was clearly overreacting to the news, and the other algorithms reacted accordingly in a bizarre herd-like manner. Clearly, better safeguards need to be put into place to prevent these sorts of shenanigans.

Advertisement

This incident brings to the mind the flash crash of May 6, 2010, when the Dow Jones Industrial average fell 1,000 points in just a few minutes.

[AFP, New Scientist, WSJ]

Advertisement

Senior staff reporter at Gizmodo specializing in astronomy, space exploration, SETI, archaeology, bioethics, animal intelligence, human enhancement, and risks posed by AI and other advanced tech.

DISCUSSION

arturo327
Arturo

High frequency stock trading is a form of rapid-fire trading that involves algorithms, or bots, that can make decisions on the order of milliseconds.

If you’re trading in milliseconds you’re going too slow. Just to give you an idea of the speeds we’re dealing with here let’s take the CME in Chicago. The main matching engine for all the CME products (bonds, currencies, agricultural products) is located in a data center in Aurora even though the market itself is in downtown Chicago. With a fiber connection out to the data center you can get a signal out there in 150 microseconds. But that’s just sending a trade from the building. High-frequency traders are operating black box servers hosted in the same data center. This cuts response time down to nanoseconds. The black boxes also use proprietary software and some form of Linux to cut lag caused by the machine itself. For stuff that’s traded out in New York like oil and gold, well, if someone’s based in Chicago the fiber connection out there is all fine and dandy but there’s also microwave transmitters atop the Sears Tower pointing to the data centers at both Aurora and Newark. This can also cut down the microseconds on your response time to the servers but it also costs something like $2 million a year to use that connection. If people are on black boxes they have a ton of money at their disposal. It’s all an arms race at this point.