Case Study: Algo Pairs at Investec
Investec Securities (UK) is a UK institutional equity house offering institutional services across research, sales, sales trading and risk trading. In September 2006, it announced that it had gone live with royalblue's Fidessa Pairs and VWAP algorithmic trading modules. AT talks to Trevor Gatfield, head of IT for Investec Securities (UK) about the group's use of the Fidessa Pairs module in its pairs trading activities.
Were you previously placing your pairs trades manually?
Yes we were. Our traders were using Fidessa to track the markets and to execute the appropriate trades.
Which markets are you pairs trading?
At present we only link to the LSE. However we are obviously very focused on the consequences of MiFID as regards best execution and so we are in direct communication with a number of clients to see which additional pools of liquidity we may need to connect to going forward. However, at present no final decisions regarding additional markets or trading venues has been taken.
Which stocks are you pairs trading - just the FTSE100?
No, our remit is wider than that. As an integrated institutional broker we have corporate finance as well as corporate broking specialising in mid-cap stocks, but we also cover larger caps as well. We are brokers to eighty companies and make markets in about one hundred and sixty stocks, with thirty-five people on the research desk in addition to sales desks and sales traders. As regards pairs trading, the new technology allows us to compete with most people and maintain our service to our clients. Obviously with mid-cap stocks you have slightly more of an execution challenge with pairs than you do for large-cap stocks.
Are you just trading outright pairs, or do you trade synthetics as well?
At present all our trades are outright pairs.
How much tolerance do you have as regards filling pairs trades? Do you insist that both legs of a transaction are completed near-instantaneously?
The system has a tolerance setting that the traders can adjust, so one leg of a pairs trade can continue to execute up to a specified level beyond which the imbalance would be deemed too great. This tolerance setting is defined on a per stock or per pair basis by the traders, whohave a certain amount of discretion within an overall framework that is set by the business.
How automated is the order process - i.e. are you using quantitative models that automate the decision to trade and automatically pass the orders to the Fidessa system for execution?
We are running models to decide which pairs to trade, but the pairs are actually input manually to the system, after which they trade automatically/algorithmically, with the majority of orders being good for the day. With the pairs vehicle we interact with our clients when we see certain interesting situations and trading will also result from those discussions.
How is your pairs trading activity split by strategy? (E.g. M & A, historical ratios or in-house models?)
As I mentioned earlier, we run models in house but I can't give a precise breakdown, though M & A would represent the smallest percentage or element.
Was the objective in taking the Fidessa Pairs module to improve execution quality?
The primary objective was to increase productivity and move away from the manual working of pairs trades, which was obviously laborious and time-consuming. We were also looking to improve the quality to our clients of best execution for pairs trading and to be certain we were executing correctly in a timely manner to the best of our ability.
What attracted Investec to the Fidessa Pairs solution in the first place?
We already had a working relationship with Fidessa that began some five or six years ago. Since then we have gone through four or five different upgrades with Fidessa to add functionality and maintain the system. Each time we have done so we have moved much closer to their core offering. (Our close involvement with them has meant that over time the core offering has also come ever closer to our exact requirements.)
Was implementation a straightforward process?
Yes, in that the Fidessa interface was already in use and so the pairs functionality was just an additional module that could be included in the next upgrade. (Upgrades are available on a quarterly basis and we can choose whether or not to take them). We have weekly meetings with royalblue, so there is therefore an ongoing dialogue that is of value when any new functionality is being implemented. Typically any user approval testing only takes around four weeks to launch and complete. Most of the work involved in these implementations is actually more concerned with testing downstream systems rather than the Fidessa module(s), though training is of course required. We find that the close relationship with royalblue means that any issues that may arise are quickly resolved and any desired enhancements usually appear in the next release.
How has the pairs module performed to date?
Extremely well, in that it is allowing us to execute business at prices that are more advantageous to our clients. We are now also able to take advantage of temporary mismatches because of the automation. In the past, we could see these mismatches but we could probably only react to one of them. Now we can deal in all of them.
The bottom line is that it has definitely given us improved quality of execution.
With the in-house pairs models might you automate the entire process end to end in the future?
This is certainly something we might consider. We will be engaging with the BlueBox product that Fidessa is launching in the first quarter of 2007. That will give us greater ability to reassess our existing pairs trading activity.
What would you say has been the main advantage of using the Fidessa Pairs module?
First and foremost, it is allowing us to manage our risks far better because we are giving the people who maintain and manage the risk and the positions more time to look at what they are doing. I'm a great believer in traders becoming execution managers, because the machine does all the basic number crunching for them. That should mean that the years of experience they have can be leveraged to best advantage - both for Investec and its clients.
Do you see the sort of automation process you have adopted as an inevitable consequence of the way in which equity markets now trade?
Yes - with automated trading books traders are doing an increasing number of trades of declining size in order to achieve the same position. (This is also evident from the growing profitability of stock exchanges.) So automation isn't a choice to us, you now simply can't exist without it.
There has been a lot of discussion over whether machines will replace human traders in capital markets. What's your view on this?
I don't think human traders are likely to disappear, but will instead be able to enjoy greater productivity because of these tools. Their knowledge of the market and their understanding of timing with regard to stopping, starting and/or combining algorithms is still absolutely crucial.
Do you think your use of trade automation will mean that your exposure to pairs trading will be increased?
I would think that this is highly likely, but obviously that will be driven by the head of trading and his traders. In a business debate I am the technology guy!