Can algos trade Trump’s tweets? Absolutely. Maybe.
Donald Trump sent the tweet heard 'round the defense industry Tuesday morning at exactly 35 seconds after 8:52 a.m. ET, blasting Boeing and suggesting he wanted to cancel the company's contract for the new Air Force One aircraft.
One second went by. Then two. No reaction on Wall Street.
It wasn't until a full 10 seconds later that Boeing stock began trading down on the news in the premarket hours, a dive that would shortly send Boeing's stock price down by as much as 1 percent in early trading, before rallying back later in the day.
The 10-second delay, which was calculated by the analysis firm Nanex, indicates that something rare was likely happening in global markets Tuesday morning: Human beings were seeing — and reacting to — news before computer trading programs could move on it.
In an era of super-fast algorithmic trading in which delays are measured in milliseconds and less, the 10-second gap indicates that possibly no one in global markets has yet figured out a way to incorporate Trump's tweets into their trading algorithms. If they had, the market response would likely have come much, much faster.
Trump's tweet said, "Boeing is building a brand new 747 Air Force One for future presidents, but costs are out of control, more than $4 billion. Cancel order!" It's the kind of remark from the president-elect that would seem tailor made for high-speed traders looking to profit from a market disruption.
"The only way that Boeing tweet could have been better is if he tagged it '$BA,'" said Zachary David, a senior analyst at KOR Group, an analytics and trading compliance consultancy.
"Twitter should start an internal hedge fund," joked Eric Hunsader of Nanex, who spotted the 10-second delay between Trump's tweet and the market's reaction.
One caveat: It is always possible that algorithmic traders are already analyzing Trump's tweets and simply decided that the tweet was too vague to trade on or that any changes wouldn't be material to Boeing.
Still, experts say there may be a reason no one appears to have created an algorithm to trade from Trump's tweets: It's very hard to do.
"I remain skeptical about the viability for such a trade because of the difficulties in determining the sentiment from a single tweet," said David. "An algorithm could have easily gotten the direction of the trade wrong today. However, given Trump's penchants for simple language and negative statements, it wouldn't surprise me if someone were able to develop an accurate predictive algorithm."
Some say they are already tackling the problem.
Efrem Hoffman, founder of a market analysis firm called Running Alpha, said Trump's tweets represent a new source of market information for those willing to study them and identify patterns. "One specific strategy that I am working on is looking at tweets that come from Trump's Android phone — as these have been shown to reflect his personal beliefs and convictions," Hoffman said. "Somewhat more unfiltered than tweets coming from other mobile devices that reflect the opinions of his colleagues/staff."
Hoffman said he is analyzing the sequencing of Trump's tweets in terms of volatility between Trump's episodes of anger or jubilation, and cross referencing those episodes with keywords associated with specific industries of policy categories. He said he is looking at the sentiment of Trump's followers and how the tweets are received as a possible measure of market player uncertainty.
"This type of indicator can be used along with the VIX curve conviction levels to identify changes in market trends on different scales," Hoffman said. One pattern he has identified so far: When market sentiment is positive, Trump's positive comments — no matter how vague — tend to get more traction in the market.
David is not convinced. "I think the estimated amount of 'Twitter algos' is vastly overblown," he said. "It's more of an academic exercise."
But, he added, the academic exercise is getting more robust every week: "If Trump continues to publicly call out individual companies via Twitter, we'll soon have an interesting data set to work with."