Gnip and Dataminr are two of the companies that sort through the Twitter API searching for relevant trade ideas. Other companies will scour through multiple media outlets and create “sentiment” analysis. Whether you’re looking at Twitter, or you’re looking at an online article from the Wall St. Journal to make an automated trade, you have to answer several questions.
Is this a rumor, or is it fact? Rumors have often moved the markets. How do you quantify them? Is this breaking news?
If it is breaking news, did someone else break it first?
Thomson Reuters pays the University of Michigan to release their Consumer Sentiment number 2 seconds before any other media outlet can release it.
If it is true, and breaking news, how relevant is it to a particular stock? News on Anne Hathaway’s box office success has often sent Berkshire Hathaway stock moving in the direction of the movie critics. Woops, wrong Hathaway..
Is this comment from a reliable source?
Well, Associated Press is considered a reliable source, but they got hacked and falsely tweeted this:
Twitter can easily be hacked. Several companies claim that they avoided the #Hash-Crash by validating the structure of the message. The hackers did not follow the specific format used by the AP for a major global or terrorist event. NewsEdge an Acquired Media company claims that their system didn’t trigger trades off of this tweet because of the structure. You’d have to ask them more about how they verified that, but their not the only company to make the claim.
The best way to structure data is to use unique identifiers. This is an important aspect to machine readable news or any breaking data that will be used for trading. The treasury has long used what is called a CUSIP number for their treasury auction results. One CUSIP will identify the auction for a particular bond from others that could be confused with it. It is a unique identifier that separates the information in a structured way. Companies like Dow Jones and Reuters will use these “tags” or unique IDs to let you know exactly what you’re looking at. They are also very useful in the world of coding. We can store and filter through large amounts of data very easily with computers if we’ve got unique identifiers for what we’re looking for. A ticker symbol would be another example of a tag identifying a specific company.
Justin Bouchard