The use of algorithms in FX trades has more than trebled in the last two years, according to researchers ClientKnowledge. The proportion of trades using them leapt from 2 per cent in 2004 to 7 per cent in 2006.
Algorithms in FX trading are still a new concept to many institutional investors. They are used quite differently in that market compared to equities, where they are more established. Justyn Trenner, CEO, ClientKnowledge, says: “With equities, a human comes up with the execution order, but you use a computer to manage the execution, to test different venues, and to time the order so that you get the best execution.
“We do see some of that in FX, but it is not the most common use. In FX, a computer dreams up the order and executes it. The humans just provide the instruction set to the computer and monitor its performance. So it is more automated.”
He adds that this 7 per cent usage is having a larger effect on the market than the figure might indicate. “Its importance is greater than its sheer volume because algo execution by the buy-side is deliberately set up to spot and punish mispricing. So [to avoid being punished] the sell-side needs to price more accurately, more quickly. There is material growth in demand for algos in FX by the sell-side to do this and manage liquidity. We are working with some organisations in that area.”
Mr Trenner predicts that take-up of algos will continue to grow in FX. However, the growth will come from use of algos that is more similar to the way they are used in equities - that is more people will want to use electronic tools to get best execution. “There is no easy money to be had in latency arbitrage, which the majority of traders in FX have used algos for to date,” he says.
Chip Lowry, senior managing director, State Street Global Link, Europe, is more sceptical. He says: “I don’t expect investment managers in FX to adopt algos as their main style of trading. It may be a tool in the toolbox. For example, some hedge funds generate some of their alpha by algo trading.
“My concern is the amount of faith people put into algorithms. Some people are taking it to the extreme where they don’t think they need a trader. But there is no algorithm for inspiration or intuition. Some of the trader’s job has to do with the feel for the market.”
Barclays Capital has been working hard on its algo offering to clients, called Powerfill. Tim Cartledge, head of e-FX trading, Barclays Capital, says: “Tools to date in e-FX have all been about executing immediately, not about being able to execute over time and manage the transaction costs. But we are starting to build that extra level of control because clients want it. They want optimal trading techniques to minimise transaction costs and to keep maximum transparency and integration with their systems. It is still early days but it is growing.”
One problem with the use of algos in FX has been the low level of basic liquidity available. How did Barclays Capital deal with that?
Mr Cartledge says: “It is all very well having an algorithm that wants to time slice an order and buy £50 of sterling-yen every 20 seconds. At that time when you want to buy it, the question is: who from? The first stage of trying to build an algo product is actually to build the liquidity.”
The first step was therefore to deepen the liquidity of Barx [Barclays Capital’s FX e-trading platform], with streaming rates. Liquidity will therefore be available for any algorithm the client wants to use. The next stage for Barclays was to build their algos from the ground up. “There wasn’t anything off the shelf. We couldn’t look to futures or equities and say we will do the same, because it is a very different market,” says Mr Cartledge.
“In FX, liquidity is also very fragmented. So we had to connect to a lot of different environments to build up our liquidity pool. It has been a problem and we spent a lot of time and resources working round it. At least now for our clients it is not a problem because we take the headache out of wiring to all the different environments. We also tune the execution algorithms on how to deal with any particular exchange at any particular time.”
At its heart, Powerfill provides the client with a simple ‘accelerator pedal’ so they can control the pace of their order. From that base the team will be able to keep building greater sophistication. Mr Cartledge says: “You can see where FX will end up by looking at other markets. Look at equities or futures – where you have volume-weighted and time-weighted strategies. Our market will have very similar offerings in the future.”
The major barrier to using algos in FX is the amount of data you need to calibrate your model. It results in a lot of number crunching and large IT and labour costs.
Mr Cartledge says: “For many people this is too hard for the benefits you get. That’s why we have identified it as our niche, as we now have the economy of scale to do it.”
Flextrade offers an algorithmic trading platform, called FlexFX, on which investors can choose which models to use.
 | Anand Iyer, vice president and co-head of FX, Flextrade, says: “The first stage of evolution in FlexFX algorithms is models which cover up basic market deficiencies by providing pseudo orders. These are order types that the market doesn’t support. |
Iyer: much less information about trades
“The second stage is the smart routing of orders which helps you find the best liquidity. Fee structures and costs are also factored into these algorithms.
“The third stage involves larger strategies that the sell-side can potentially support. The problem with this stage is that the FX market is over the counter, so there is a lack of benchmarks. In equities, everything traded gets printed to tape. In FX it doesn’t, so there is much less information available about the trades, for example volumes. It means you have to guess, using whatever you have. So the algorithms take a different shape.
A good example of these strategies is the previously mentioned time-slicer, where you chop your trade into smaller amounts to avoid moving the market. Mr Iyer says a fair number of his clients now use this strategy.
 | Mark Palmer, vice president and general manager, Apama products, Progress Software, says that growth in the use of algos in FX has been very quick and sudden over the last year. He says: “The main reason for that is the nature of the markets themselves. The liquidity pools are so fractured and they have such different characteristics. Wherever there is that uncertainty, it just creates more opportunity to arbitrage and do interesting things and find alpha. In comparison with equities, the FX market is much more innovative.” |
Progress’ Apama Algorithmic Trading Platform uses what it calls complex event processing (CEP) technology. This navigates volumes of data – including FX data - to automate trade execution. The data is used to develop, deploy and back test trading strategies quickly, before the market opportunity passes.
Mr Palmer says that the crucial element of CEP is that it is dynamic and can re-evaluate rules and act according to changes in the market.
“The market changes so fast that the algos will tell the analytics to adapt in order to compete, says Mr Palmer.” That is the whole focus for our platform: to be able to create and change these things rapidly.”
Just as the markets change constantly so do the barriers to take-up of algorithmic strategies. Mr Palmer says: “Six months ago the barriers were about getting good liquidity, training people and discovering algos. We have changed a lot now. Most firms now have a basic understanding of algos. They know they have to do it because the competition is doing it.
“Now the biggest questions are: ‘How do I evolve quickly? How do I come up with new models and figure out which ones are the best? How do I know when someone is front running me?’ Algorithms fighting algorithms is where the battle is starting now.”
Mind blowing decisions
So you want to use algorithmic trading strategies in FX? How much investment do you want to put in? Do you want to hire a team of stats PhDs and an IT team to build your own model? Or do you want to partner with a bank or other provider?
If you want a new piece of proprietary research and technology you would probably want to build your own. Many hedge funds are doing just that. But if you want to minimise your costs, the best approach would be to choose an existing algorithm provider.
Once you have chosen a provider, the next thing is to choose which algorithms you want to use. This is probably the hardest part.
Mr Cartledge says: “At the lowest level, you are figuring out when and at what level to put orders into exchanges. I can’t tell you the secret formula because there isn’t one, it changes constantly. What works today, won’t work next month because the market is dynamic. The more people do one thing, the less advantage it has. You have to look at a lot of data continually and keep choosing the low level algorithms and how they interact with the rest of the world.
“At the higher level, it is much more constant. What degree of risk are you comfortable with for your order? If you want to buy $100m euros as euro-dollar, what time period are you happy with? Five seconds is different from 30 seconds, which is different from a minute.”
Mr Palmer at Progress Software says it is essential to test your new ideas for algorithms to find out which ones work best. Apama has developed a system it calls genetic tuning - to weed out all the bad algorithms by testing them on historical data. He says: “They say that the little toe isn’t needed, eventually we will all lose them in a billion years. In algorithms, the evolution doesn’t take a billion years - it takes a week.”
Spot FX trading system, EBS, also offers the ability to test algorithmic models on historical data. Investors can use the system’s lab to test connectivity, timing and price integrity with actual spot market data to analyse program performance, of a particular trading programme.
Jack Jeffery, CEO, Icap electronic broking, says: “The key to successful algorithmic trading is to have access to the best liquidity, pricing and diversity of counterparties, underpinned by strong and proven technology, which is what EBS provides. Commodity trading advisers and hedge funds need the facility to test their algorithms with real market data before applying them to a live environment to ensure the maintenance of an orderly market.”
