I recently had the opportunity to meet with the CEO of a fascinating technology company called iSentium. The company has the ability to analyze some 50 million Twitter messages per hour to search for signals about the stock market’s reaction to social media messages.
One of the interesting anecdotes from that meeting was the observation that an estimated 97 percent of news headlines make their way onto Twitter before they appear on the major news wires. Twitter’s impressive propagation speed – the time it takes for a message to make its way through webs of interconnected accounts via “Retweets” – has drawn the interest of data analytics firms.
For companies engaged in high-frequency trading (HFT), access to data distilled from Twitter and other social media sources has become increasingly important. The marriage between super-fast purveyors of data and super-fast traders of stocks that react to new information is such a good fit, its occurrence seems almost inevitable.
While there are many types of high-frequency traders, the media has fixated on a particular aspect of their operations: latency arbitrage. Latency refers to the delay between a trading decision and its execution. In his book, Dark Pools: The Rise of the Machine Traders and the Rigging of the U.S. Stock Market, author Scott Patterson provides a fascinating description of another HFT phenomenon: co-location, the housing of trading computers in the same facility as an exchange for the purpose of optimizing speed.
In a recent blog article, financial journalist Felix Salmon observed that HFT is much more complex than just exploiting the advantages that stem from latency arbitrage and co-location. The process of latency arbitrage – which involves, as he puts it, “trying to buy or sell at yesterday’s prices, in the knowledge of where the price is today” – is just part of the picture.
While it’s clear that there’s much more to HFT than latency arbitrage, it’s hard not to be impressed by the potential for some trading institutions to realize an economic advantage from the lightning-fast speed at which Twitter processes data. The speed at which news events are disseminated can be an advantage for firms which can execute trades faster than others, even if the time differential seems infinitesimally small to non-professional traders. In the world of Twitter data propagation, time is measured in milliseconds (i.e., 1/1000 of a second).
An academic study of the propagation speed and social influence of Twitter analyzed the millions of Tweet replies that had resulted from a sample of 58 million Twitter messages. The study revealed that 37 percent of message flows had spread more then “three hops” away from the message’s originator. According to the study, 25 percent of those replies were generated within 67 seconds of the original tweet.
There’s a clear convergence underway among big data, social media analytics and institutional trading. Where this convergence will ultimately lead, and what its implications will be for individual traders, will be the subject of ongoing debate. Whatever the outcome of that debate, social media analytics are likely to play an increasing role in the world of institutional trading.
What about non-professional traders? Some investors — those individuals who hadn’t received wedding invitations to the union of “fast Tweets” and “fast trades” — may inevitably find themselves using “yesterday’s news” to trade stocks at “yesterday’s prices.”