XML data for algorithmic trading March 8, 2007
Posted by James Webster in : finance , trackbackFinextra reports on Dow Jones XML news feed for building trading algorithms that react to breaking news stories.
Hmmm… I wonder if anyone is game enough to plug the RSS feeds of Digg, Slashdot, TechCrunch or Valleywag into their algo in their attempt to gain alpha?
Comments»
Cool stuff. I always thought Google could be a force in this space with all of the information they have. Maybe someone there will take advantage of this (and Google’s huge data store) on their 20% time.
People have been doing this for years with varying success. The Dow Jones feed isn’t anything new, other than their use of XML. At one point in the XML craziness, people were suggesting a FIXML format.
Putting something into XML doesn’t make it inherently more useful, you still have to do something with the data.
Reuters has offered a news feed for a long time. With each news item, it includes a set of RICs which you can index the article by. You can then try to correlate price changes to news, or Bayesian score words in the news article.
I’ve seen programmers parse the Reuters feed using Visual Basic, and do some cool stuff. Give an XML feed to a Java programmer and first they’ll spend months talking about DOM/SAX, Schema vs DTD, character encoding, etc.. Then they’ll feel they need to use JAXB or something to translate the XML into Java classes. But the generated Java classes shouldn’t be used directly, so they’ll create another layer on top of it. Soon, they’re think about what O/R tool to use to put the objects in a database. They’ll create such a mess of classes and interfaces, that they’ll need Spring to keep everything held together.
By this time the VB programmer has delivered a working product to the business and moved on to other projects.
One of the more interesting news feeds I’ve seen came from a company which simply screen scrapped various corporate news websites. When companies make announcements, they attempt to release it to all media sources simultaneously, however, it’s been found that quite often the news shows up, or can be found on their website before it’s released to the general public. A company with using a CMS may create an article on their website, and not link to it until moments before the news announcement. A knowledgeable person can often guess URLs for content which hasn’t been publicly linked to yet. Getting news just a minute or so early, can be a large advantage to a hedge fund.
Yep, you can build a profitable business trading on news, but you have a more profitable business if you are the one selling the news.