Before the written word, information flow was severely limited. Whatever, wasn’t in human memory was basically lost. To use a computing parallel, it was as if everything was in volatile memory. By constant, having the written word, meant we could store words on long term storage mediums, such as stone, later paper and now digitally. The cost of this long term storage has consistently gone down over the time. The problem is that in the modern day there is simply too much written information for any human to read, maybe a computer could help us read the written word, and particular to give us a better feeling of what the “news” is?
The notion of using news, is certainly not new within the realms of trading. It has been going on for centuries. Traders now consume news from many different sources, in particular newswires such as Bloomberg News and also increasingly from other sources like social media, to aid them to make trading decisions. The challenge of course is filtering all the noise to find usable information. This is especially difficult, given there is so much news being generated these days, it is impossible for a human to read it all. If you are covering many assets, it is impossible to know what is going on everywhere. A computer by contrast can parse thousands of news articles very quickly!
However, not as many people as you would expect are using machine readable news to help improve their trading decisions. This is despite the fact that machine readable news has been around for over a decade. For example, I remember seeing a presentation discussing machine readable news when I was doing currency research at Lehman Brothers. The major newswires offer machine readable products to allow computers to parse their news in a convenient format.
Why is machine readable news still at a relatively early stage in terms of take up? I think there are a lot of misconceptions about machine readable news in finance. After several researching using machine readable news to trade markets, I thought it would be worthwhile to clear up a lot of these misconceptions! Here are a few points below which I think are worth knowing if you want to start delving into machine readable news for trading. It is an exciting area, and I think in the coming years, take up of machine readable news is likely to grow significantly.
You can use machine readable news to help discretionary trading: it’s not just HFT trading headlines
There is an image of machine readable news all being about speed, making split second decisions. Yes, there are high frequency traders who systematically trade news headlines. However, this is certainly not the only use for machine readable news. It isn’t all about latency arbitrage. Indeed, when I’ve done research such as for Bloomberg, using their machine readable news (get the paper here), my goal was to aggregate many news articles into a daily sentiment indices, that could be used both by systematic traders and discretionary traders. It can be used to trade markets at longer time horizons.
You don’t need to do everything from scratch: using structured news dataset will save time
It can be quite time consuming to read news articles in their raw form from the web. Furthermore, you don’t have access to the bulk of news articles from newswires, which traders typically read on their desktops (and should be the ones you look at too!). Newswires like Bloomberg offer their news in a machine readable form, which is nicely structured for computers to read. They include a wealth of metadata to help make sense of it, such as what is the topic of the news, what tickers are associated with it etc. Very often the full text of the article is also included if you want to do further analysis on the text, such as running some sentiment analysis applications. There’s enough data there to come with your own strategies and indices, whilst at the same time avoiding a lot of the more time consuming data cleaning steps.
Machine readable news isn’t just for equities: Use it for FX too!
One of the most prevalent uses for machine readable news is for trading single stock equities. It is understandable because it tends to be easier to associate news articles for equity entities. Even then there are difficulties such as being able to map brands to specific equities. By comparison, machine readable news isn’t used by as many folks in currency markets. One approach to using news in FX, is to filter by currency specific articles, and I’ve used this approach. Another way, is using your knowledge of currency markets to filter other types of news, such as economic news. I’ve also researched this in the past. As another example, I’ve also used news associated with FOMC and ECB meetings to model FX volatility.
Machine readable news doesn’t need to be used on its own: use it as an additional factor
One of the advantages of using machine readable news from newswires is that the content is written in a consistent style (and there is generally a lot of content too). This contrasts with social media which by it’s nature is more challenging to understand for a computer, given it’s written in an inconsistent way and consists of very short messages. However, in practice, we can news as a factor alongside social media, and indeed all the traditional factors we look at, such as trend and value. Social media can still provide valuable signals, but we need to treat it differently to news data. As an example, I have developed forecasts for nonfarm payrolls using social media as one of the factors.
Interested in using machine readable news to trade? Who should you ask? Cuemacro, of course!
Cuemacro has experience working with many unusual datasets, including machine readable news (including Bloomberg News) as well as datasets from social media (including Twitter). Drop us a message if you would like to use our consulting services to see how you can use machine readable news to help you make better decisions in markets, whether you’re a discretionary or systematic trader.