Don’t make a hash of it!

Last week’s fake tweet set US markets into an instant free fall and showed thatmarkets still react to misinformation as quickly as verified fact.

By None

Last week’s fake tweet set US markets into an instant free fall and showed that markets still react to misinformation as quickly as verified fact. In an age of highly automated trading, the ‘hash crash’ has forced an examination of how news should be input into trading strategies.

Shortly after 1pm on Tuesday 23 April, a tweet from the verified Twitter account of US newswire Associated Press stated explosions at the White House had injured President Obama. Within minutes, the markets had bottomed out, with the Dow Jones Industrial Average sliding 145 points, or 1%, before rebalancing pre-drop, four minutes later.

Since Tuesday, the Federal Bureau of Investigation and Interpol have formally begun efforts to find out who was responsible, with an organisation named the Syrian Electronic Army publicly claiming to have hacked the AP’s Twitter account and posted the fake message.

The automation of news and information to fuel trading strategies started with programmes that read economic data releases but the activity has broadened to encompass a range of tools that scan breaking information and feed relevant data into electronic trading strategies.  The growing use of social media as a distribution mechanism for potentially market sensitive information has focused attention on how such inputs are or should be factored into trading decisions.

“Given the exposure the tweet had across a number of platforms and millions of followers, it’s likely the downward spiral the market was attributable to the human response to the ‘news’, not necessarily to an algo responding automatically to a machine-read tweet,” Richard Brown, global head Elecktron Analytics for Thomson Reuters, told

Filtering process

Brown, who runs Thomson Reuters’ automated news feed business, believes the downward spiral and subsequent resurgence was accelerated by the prevalence of electronic trading, which let the market quickly recover after the tweet was proven false.

Automated news reading services scan web-based news sources and social media for breaking news relevant to markets and specific securities. This can be fed via an application program interface (API), to a trading algorithm, which may act as a circuit breaker to stop trading or accelerate participation in a certain stock depending on the news.

Joe Gits, president of Social Market Analytics, a firm that feeds social media information into trading algos, said market participants must not forget the nature of social media – that of unverified discussion, rather than a traditional news outlet.