Will Human Traders Be First To React To News Events Ever Again?

by Hugh Taggart
  • Depending on how you look at it, of sharing an office with two discretionary forex traders is a double-edged sword.
Will Human Traders Be First To React To News Events Ever Again?
Photo: Bloomberg
Join our Telegram channel

I have the fortune, or misfortune - depending on how you look at it, of sharing an office with two discretionary Forex traders who manage a bit of money for wealthy individuals. They are forever moaning, at data time, how the market moves before the data are available to the general public. How can this happen – legally?

Well, it isn’t legal, I hear you say... And the current investigations into daily fixings of FX prices and gold, the recent LIBOR scandal, and numerous other well-documented cases, would lead the average person to believe that under-the-counter methods are being used to get data early. Possibly, or even probably, true. But not much to say here apart from we hope the regulators get through those investigations soon.

There are ways, however, certain traders have a head start on both expected events - like data releases, and unexpected events - such as terrorist attacks or key speeches. And this is where the subject of Machine Readable News (MRN), aka news Analytics , comes in.

There are two main ways a trading organization could legally be one of the first to react to an economic or geopolitical event. The first is by using a low latency economic data feed. The second is by using systematic event detection from text. I’ll use the rest of this post to elaborate on these two methods a bit…

Vendors of low latency economic data are commonplace now. To name a few, Dow Jones Elementized News Feed, Thomson Reuters News Feed Direct, AlphaFlash, Bloomberg Event Driven Feeds and NASDAQ OMX Event-Driven Analytics.

These firms compete to deliver economic data from the ‘lockups’ – the room where journalists are invited to publish economic data at the correct time – to clients with the lowest possible latency, i.e., the quickest. It’s an arms race for the vendors – who has the fastest networks, the lowest number of servers and so on. The client, in turn, has a computer set up to receive the data and trade on it within milliseconds of receipt. Hence we’re talking response times in tens of milliseconds after the data is published. To put it in context, it takes us a few hundred milliseconds to blink.

Then there are firms who react quickly to unexpected events – and this is probably where the most value lies in the future. These trading firms use some form of text mining technology to detect events on newswires, on the Web or, as some people would lead us to believe, on Twitter. More on Twitter in a second, but this technology is improving all the time and many firms now use it to systematically react to anything from a natural disaster or act of terrorism, to comments by central bankers.

How does the technology work? It’s rules-based, mainly. Computers are programmed to scan text for patterns of language that match certain event types. The events are specific to an entity – a place, an organization – like the Fed or ECB, or tradable instrument. The data provided also include a sentiment score that means the client can trade directionally, e.g. earthquakes are negative for an economy, higher unemployment guidance is negative, higher GDP guidance is positive and so on. And all of this takes a couple of hundred milliseconds – way quicker than a human would normally react to an unexpected event.

So that’s why, if you’re a discretionary trader, you’ll get beaten to the punch every time… You’re left to trade the over-reaction – or reversal – and the longer-term implications of each event.

Now briefly to Twitter. There are firms trying to systematically trade events based on Tweets, but it’s extraordinarily difficult because of the amount of noise being published., meaning it’s very easy to get it wrong and lose your shirt. So it’s unlikely anyone is successfully using Twitter in this way yet. This has been confirmed by many people who use or produce MRN. The most public case of Twitter’s alleged role in MRN was when the AP’s Twitter account was hacked back in April. Some said traders using Twitter systematically where behind the so-called ‘hash crash’. But data from Nanex https://www.cnbc.com/id/100669831 said it took 17 seconds for the market to react to this news – that’s not a reaction led by machines. Exacerbated, for sure.

I have the fortune, or misfortune - depending on how you look at it, of sharing an office with two discretionary Forex traders who manage a bit of money for wealthy individuals. They are forever moaning, at data time, how the market moves before the data are available to the general public. How can this happen – legally?

Well, it isn’t legal, I hear you say... And the current investigations into daily fixings of FX prices and gold, the recent LIBOR scandal, and numerous other well-documented cases, would lead the average person to believe that under-the-counter methods are being used to get data early. Possibly, or even probably, true. But not much to say here apart from we hope the regulators get through those investigations soon.

There are ways, however, certain traders have a head start on both expected events - like data releases, and unexpected events - such as terrorist attacks or key speeches. And this is where the subject of Machine Readable News (MRN), aka news Analytics , comes in.

There are two main ways a trading organization could legally be one of the first to react to an economic or geopolitical event. The first is by using a low latency economic data feed. The second is by using systematic event detection from text. I’ll use the rest of this post to elaborate on these two methods a bit…

Vendors of low latency economic data are commonplace now. To name a few, Dow Jones Elementized News Feed, Thomson Reuters News Feed Direct, AlphaFlash, Bloomberg Event Driven Feeds and NASDAQ OMX Event-Driven Analytics.

These firms compete to deliver economic data from the ‘lockups’ – the room where journalists are invited to publish economic data at the correct time – to clients with the lowest possible latency, i.e., the quickest. It’s an arms race for the vendors – who has the fastest networks, the lowest number of servers and so on. The client, in turn, has a computer set up to receive the data and trade on it within milliseconds of receipt. Hence we’re talking response times in tens of milliseconds after the data is published. To put it in context, it takes us a few hundred milliseconds to blink.

Then there are firms who react quickly to unexpected events – and this is probably where the most value lies in the future. These trading firms use some form of text mining technology to detect events on newswires, on the Web or, as some people would lead us to believe, on Twitter. More on Twitter in a second, but this technology is improving all the time and many firms now use it to systematically react to anything from a natural disaster or act of terrorism, to comments by central bankers.

How does the technology work? It’s rules-based, mainly. Computers are programmed to scan text for patterns of language that match certain event types. The events are specific to an entity – a place, an organization – like the Fed or ECB, or tradable instrument. The data provided also include a sentiment score that means the client can trade directionally, e.g. earthquakes are negative for an economy, higher unemployment guidance is negative, higher GDP guidance is positive and so on. And all of this takes a couple of hundred milliseconds – way quicker than a human would normally react to an unexpected event.

So that’s why, if you’re a discretionary trader, you’ll get beaten to the punch every time… You’re left to trade the over-reaction – or reversal – and the longer-term implications of each event.

Now briefly to Twitter. There are firms trying to systematically trade events based on Tweets, but it’s extraordinarily difficult because of the amount of noise being published., meaning it’s very easy to get it wrong and lose your shirt. So it’s unlikely anyone is successfully using Twitter in this way yet. This has been confirmed by many people who use or produce MRN. The most public case of Twitter’s alleged role in MRN was when the AP’s Twitter account was hacked back in April. Some said traders using Twitter systematically where behind the so-called ‘hash crash’. But data from Nanex https://www.cnbc.com/id/100669831 said it took 17 seconds for the market to react to this news – that’s not a reaction led by machines. Exacerbated, for sure.

!"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|} !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}