Mining Twitter for Trading Signals

Start-Up Profile: Social Market Analytics

The explosion in social media has created new ways to measure investor sentiment and turn that information into actionable intelligence about market direction. With this article, MarketVoice launches a series spotlighting companies with innovative financial technology.

ON MARCH 27, at a little after 3:30 p.m. East Coast time, a Wall Street Journal reporter named Dana Mattioli flashed out a headline announcing a scoop. “Intel is in talks to buy Altera,” she wrote, adding that details would be coming soon on the paper’s website. Altera’s stock price quickly began to rise, so much so that trading in its shares was halted only three minutes later. At the closing bell, Altera’s price was $44.39 per share, a gain of 28% on the day.

How did the news get out so quickly? One set of investors got the headline via the Dow Jones newswire, which provides real-time news and information to thousands of subscribers worldwide. But there was another way to get that information almost as quickly. Like many reporters nowadays, Mattioli posted a note on Twitter announcing the scoop almost immediately after the headline went out on the wires, making it instantly available to anyone with an internet connection and a Twitter account.

It is difficult to know for sure how many people saw her tweet and acted on it during that brief three-minute window. But this incident shows why many traders are now looking to Twitter as an additional source of market-moving information.

500 Million Per Day

Every day millions of people post information on Twitter, and a rapidly growing number of those messages are related to financial markets. Traders share ideas about what they are buying and selling, analysts offer insights on market behavior and predictions about future price movements, and occasionally reporters reveal marketmoving news on their Twitter feeds.

Twitter also has become a venue for prominent investors to share their views. On April 21, Janus Capital used the online network to share a trading idea from Bill Gross, the renowned fixed income investor. “German 10 yr Bunds = The short of a lifetime,” Gross said in a message posted on Twitter. The fact that he used Twitter to share his ultra bearish views with the world reveals how much the network has become part of the financial market landscape.

All of these messages are public, which opens up new possibilities to gather market-moving information and identify changes in investor sentiment. The problem is sorting the wheat from the chaff. There are now more than 280 million active Twitter accounts and more than 500 million tweets posted per day. No human being could possibly scan all those tweets looking for the handful of messages with useful information.

That’s where companies like ours step in. We are part of a wave of firms that collect, sort and filter messages on social media and use that information for a wide range of analytical purposes. Within that general trend, there are a small but growing number of companies that focus on messages pertaining to traded assets and measure the sentiment expressed in those conversations. Our business models are based on turning that sentiment into actionable intelligence about the future direction of prices.

The SMA Business Model

At its most basic level, analyzing social media messages is basically a new way to get the wisdom of the crowd. SMA scans the stream of Twitter messages looking for mentions of companies, tickers, products, commodities, currencies and anything else that relates to a traded asset. SMA hones in on messages that express the intentions of a key subset—active traders and other market professionals who frequently comment on market drivers and traded assets.

In general there are three types of messages that SMA captures. Some messages contain important news, such as the tweet about Altera mentioned above. Those don’t just come from reporters, by the way. A farmer who tweets about the quality of his crop or an oil company employee who tweets about a refinery fire is providing valuable information.

Another type is analysis. There are scores of analysts who use Twitter to broadcast their observations and insights. Many of these analysts do not work for large well-known institutions such as investment banks, but as companies like Estimize have shown, their track record is as good as if not better than traditional sell-side research.

The third type is opinion. A huge community of traders has sprung up in the last five years or so who use Twitter to share opinions about the market and ideas for profitable trading. The majority of these traders are individuals who trade for their own account, but there are lots of big names as well. One of the best examples is Carl Icahn, the billionaire investor and activist shareholder.

Social media conversations are similar to the voice of the crowd on a trading floor. Just as the level of noise can tell a veteran trader something important about the mood of the crowd and the quality of the market, the volume of social media messages can send an important signal about the collective mood of the trading community.

In August 2013 he sent out a tweet about his views on Apple’s stock that caused the company’s value to increase by $17 billion in about one hour.

The New Trading Pit

One of the biggest concentrations of finance-related social media conversations is on StockTwits, a platform geared for people interested in trading, investing and markets.

StockTwits estimates that more than 300,000 people share information and ideas on its platform, producing streams of messages that are viewed by more than 40 million people across the internet and through other social media.

Like Twitter, a message on Stock- Twits can only be 140 characters. Also like Twitter, users can identify the asset they are discussing by using the dollar symbol followed by the ticker symbol. For example, $GOOG tells the user community that a message is about shares in Google. That makes it easier for users to search for messages and follow conversations about that stock. The use of the dollar sign was first used for asset identification on the StockTwits platform in 2009 and was adopted later by Twitter, where it is commonly referred to as a “cash tag.”

As the name says, StockTwits was centered on stocks when it started in 2008, but an increasing number of messages on its platform link to futures, commodities, currencies and other traded assets. Cash tags combined with futures symbols are used to identify specific contracts. For example, the crude oil futures traded on CME Group areidentified as $CL_F, and the E-mini S&P 500 futures as $ES_F.

In a sense, these social media conversations are similar to the voice of the crowd on a trading floor. Just as the level of noise can tell a veteran trader something important about the mood of the crowd and the quality of the market, the volume of social media messages can send an important signal about the collective mood of the trading community. But there is a lot more to social media analytics than simply measuring the number of positive or negative tweets.

Filtering and Scoring

Social media is like every means of communication; there are spammers, scammers and con artists. SMA has developed algorithms to filter out messages that are deliberately intended to create false impressions. For example, someone who tweets sporadically and then sends many tweets in succession is more likely to be attempting manipulation. Messages that are irrelevant are also filtered out. Someone who talks about an Olympic gold medal probably is not talking about the price of gold.

The next step is to identify tweets that express expectations. A tweet that mentions selling $CL_F two weeks ago does not help predict crude oil prices today. SMA uses a natural language processor to identify forward-looking tweets with phrases like “going long,” “buying puts,” “raising rating,” “hitting support” and “broke through resistance” that are representations of trader expectations.

One of the challenges is to aggregate the specialized terms that exist in these markets. Tracking social media conversations about the exchange rate for the British pound against the U.S. dollar means you have to search for not only $USDGBP but also “cable,” which is trader jargon for that currency pair. It’s also important to recognize relationships among similar products, such as exchange-traded funds that track the price of commodity futures.