The Gateway to Algorithmic and Automated Trading

New News

Published in Automated Trader Magazine Issue 23 Q4 2011

Machine readable news has been with us for a while now and has moved into the mainstream as an input to traders' alpha and execution models. To date, the majority of services have primarily focused on major economic announcements, but traders are already looking for more from providers in terms of both flexibility and new data. Georg Gross, Head of Front Office Data & Analytics, Deutsche Börse AG, examines some of these new areas of demand and opportunity.

Georg Gross

Georg Gross

From being a novelty headline just a few years ago, machine readable news has become a practical reality for an increasing number of traders. About 15% of respondents to Automated Trader's 2010 reader survey stated that they involved event or machine readable news feed data in their trading algorithms. That number had risen to more than 19% in the 2011 survey. In addition, nearly 34% of respondents anticipated using machine readable news in this way within the next two years.

It goes without saying that for traders such as these, speed of delivery is paramount; the window of profitable trading opportunity in the aftermath of a major economic announcement is small. Therefore, the provider who can deliver the news to its trading clients first is adding real value to the alpha capture process. This isn't just a simple matter of wire speed: the point of delivery also matters. In an ideal world, the data should be delivered to major co-location facilities that are as close as possible to either the source of the information or the relevant trading engines (or if possible, both).

To achieve (and keep reducing) the minimum latency between the source and the clients' trading engines, a provider of machine readable news needs to manage the entire delivery chain - starting from the lock-up rooms (or other source) through to the clients, via the lowest latency lines. When one considers that for some news events the difference between first and third place in terms of delivering the news may be less than a millisecond, this emphasis on end-to-end oversight over latency reduction is essential. This is the approach that Deutsche Börse takes with its AlphaFlash® machine readable news feed, to ensure that for the majority of news events AlphaFlash® data arrives first.

Another critical factor is obviously the quality of the information that is being provided. The importance (in terms of its ability to move markets) of the machine readable news being delivered is obviously paramount as well. Including generic 'filler' news in the feed is pointless and ultimately disruptive from a client perspective; the feed should only include the most price sensitive events that have genuinely actionable information value for clients' trading models. In practice, that has meant that machine readable news has historically been confined to major economic announcements, but as participants have developed greater insight into the quantitative relationship between news events and price action, their requirements have started to expand.


As client expectations of the content range of machine readable news have started expand, this has placed a greater emphasis on providers' flexibility and service quality. It is increasingly common for users to request that a specific new economic report be added to the feed, even if this event might only be of market moving significance on one occasion. Responding to requests of this nature requires technology that is sufficiently flexible to quickly incorporate the new event. It also requires a combination of both substantial resources and an entrepreneurial mindset, which are often seen as mutually exclusive.

If the provider can offer that combination of qualities, then it should be possible to turn around these one-off client requests virtually overnight, although it is generally advisable to allow sufficient time for the client to program to the new data.

This combination of flexibility and close collaboration with clients can also be used to extend the type of events reported by machine readable news feeds. Although the majority of client requests that Deutsche Börse's Market Data & Analytics segment receives relate to additional economic events, other types of events are also feasible.

For example, the Chairman of the Federal Reserve's speech at Jackson Hole, Wyoming last August was widely expected to include mention of a third round of quantitative easing. The speech was to be delivered on Friday; on Wednesday of the same week, the AlphaFlash team received requests from different clients requesting that salient points from the speech be transmitted to them as a series of AlphaFlash® machine readable news messages.

The AlphaFlash team spent several hours working with clients determining precisely which quantitative items in the speech (or any items that could be represented as a binary yes/no) would be transmitted. For example:

• Is a third round of quantitative easing mentioned?

• If so, is a specific amount of money mentioned, and how much?

Once the specific items were agreed, a message was sent to all clients advising them of the availability of the new data. Then, as soon as the embargo expired and the Chairman of the Federal Reserve started speaking, the machine readable news containing the agreed data was transmitted to clients as an AlphaFlash® message. The entire process, from receiving the first client request, to encoding the data, to clients actually trading on the news was completed within 36 hours.

Depth and breadth

In addition to a willingness and ability to collaborate effectively at short notice with clients, a provider of machine readable news needs to be systematically extending the depth and breadth of its offering as a matter of everyday policy. This may involve digging a little deeper into mainstream market events. Deutsche Börse's new Global Treasury Auctions product is a case in point. Launched in September, the new feed disseminates about a dozen auction results, such as whether there is a high bid to cover ratio or how over/under subscribed particular bond issues are.

AlphaFlash Global Treasury Auctions includes key treasury auction data from countries like the UK, Germany, Australia and Canada (the U.S. has been offered since April 2010 when AlphaFlash first launched). Other countries will follow later. The intention is that clients will be able to start making automated trades immediately based upon this data.

Unscheduled events

Most of the emphasis among providers of machine readable news has to date been on scheduled announcements. While this is valuable in its own right, it still leaves automated models at a distinct disadvantage to human traders as regards unscheduled announcements that have a price impact on the market. These could be announcements that are scheduled for a particular day, but no specific time, such as a quarterly earnings update. Or they could be an unexpected event that the company is obliged to announce by law because it believes it will have a material effect on its share price.

In these situations, trading models typically respond to the price/volume changes created by human traders' reactions to the unscheduled announcement. The timeline for this obviously varies, but it typically takes between seven and 10 seconds for human traders' reactions to be reflected in the market price/volume data. As a result, the automated trading models are having to play catch up from the outset.

This situation has been a major factor behind the development of Deutsche Börse's recently launched AlphaFlash® Corporate News Germany product. This covers some 100 of the most liquid German stocks and incorporates technology capable of recognising the sort of ad hoc disclosures (such as an earnings warning or change in dividend) that will have a price impact. The quantitative data from the announcement (the nature of which is pre-defined, so client trading models will know how to process it) is then automatically extracted and sent to clients via an AlphaFlash® message.

Even after allowing for processing time and certain necessary quality assurance procedures, this still puts automated models in a far better situation than the current status quo. They will typically be able to trade on the AlphaFlash® message content within the first second of the original announcement. In effect, they are now truly able to trade upon the news itself, rather than the initial human reaction to it.


The type of organisations typically using machine readable news as inputs to their models are pushing the trading envelope and therefore demanding in their expectations. Those expectations will obviously include a provider's ability to deliver the most extensive range of price sensitive news data faster than its competitors, but that alone is not enough. In addition, this type of organisation is looking for a service that is well-resourced yet entrepreneurial and can deliver innovative collaboration at short notice.

That's AlphaFlash® - an innovative service that offers algorithmic news.