Growing pains
The growing use of FX algorithms is throwing up new technological challenges for the FX market in terms of smart order routing and latency, reports Frances Maguire.
Now that FX is being increasingly traded as an asset class, the need for rapid algorithms for FX has become more important, but the latency issues facing the FX market, where co-location is not really an option, are very different from those of the equities market. Much of the latency exists in the brokers’ internal legacy systems, and as a result smart order routing is being used to source not only the best price but the best fills.
Kevin Covington, head of business strategy at BT Global Financial Services, says that the virtual nature of FX makes it harder to see where business is being done. “Unlike equities, with exchanges and physical venues, it is much easier to change business processes because there are clear cut technology demands and networks.”
At the same time, Covington says that the fragmentation of the number of FX providers and connections they provide mean that the growth of the FX business on BT’s network is a clear indicator that algorithms are most likely being used to access pools of liquidity.
The significant numbers of FX players provide distribution and reach using BT’s shared infrastructure and network, the same model of co-location has not evolved in FX, and most cannot be emulated. Furthermore, in the equities market the problem of improving latency is moving increasingly onto the laps of the order routing providers because they have a closed infrastructure, and no matter how much the exchanges and networks lower latency, trades can still get slowed down by firms’ internal technology. Covington says the same problem is occurring in the FX market. “As most of FX business is done over provided platforms, BT is limited in the amount it can dynamically influence performance and latency. It is not a simple co-location, fast network play,” he says.
Giles Nelson, director of technology at Apama, says that there is increasing evidence that algorithms are being used in the foreign exchange market. He says: “Buy-side firms that are accessing multiple banks, and have a significant FX trade to execute, will use an algorithm to gain an aggregated view of the market and determine the best price. Furthermore, they may also use algorithms to ‘iceberg’ large orders to execute parts of the order over time with the bank offering the best rate, echoing very closely the use of algorithms in the equities market.”
Nelson adds that the market for algorithms is split into two parts: the mainstream execution-only algorithms and the signal-generation algorithms, which look for particular sequences of events to trigger a trade. He says a quantative or statistical arbitrage hedge fund is most likely to use both of these types of algorithms in FX, whereas a traditional fund manager is more likely to use execution-only algorithm to gain greater efficiency.
As fragmentation within the FX market may increase, the need for market aggregation to get a single view of the market will become more critical. Nelson says: “By having technology to allow market aggregation, it enables smart order routing to direct orders to the party offering the best price. This is what it means to be able to intelligently deal with fragmented liquidity.”
Apama provides a core platform, which uses Complex Event Processing (CEP) technology to take in and analyse trading and quote data, connectivity, both to portals and to banks using the FIX protocol, and solution accelerators for market aggregation, as well as out-of-the-box algorithms to get customers started. These can then be customised or replaced by internal algorithms. “A more progressive hedge fund or proprietary trading shop will probably want to take ownership of the algorithms they are using, in a bid to beat the market and find alpha,” says Nelson.
The impact of latency is slightly different in the FX market, where co-location is not possible. However, even with market aggregation, it is possible that the streaming prices from the banks are hitting the user’s system at slightly different times, so an exact snapshot of the best price can be taken. Nelson says: “This is one of the reasons why the buy-side uses algorithms to split their orders up into smaller parts. They don’t necessarily want to put a large order into one particular bank. This is almost like a VWAP in equities, where the FX market is trying to benchmark against the market.”
He says that Apama has seen just as much growth in FX as the other asset classes over the last 18 months, propelled by the fact that electronic trading in FX has escalated over the past few years. “In a way, it has been easier for firms to embrace algorithmic trading in FX because there are not the same legacy systems as in the equities market. But although the uptake has been quick the market has yet to mature,” Nelson adds.
Vijay Kedia, president and CEO of US-based vendor FlexTrade, says the FX market is likely to remain fragmented and the lack of transparency in the FX market means that the trend towards algorithmic trading will continue. He says: “The benefits of algorithms are most obvious in a very fragmented marketplace. Once the buy-side has got a taste of algorithmic trading, and the benefits of it, there will be no turning back. More and more banks are now gearing up to provide streaming quotes, which they were clearly not doing 12 months ago, so there will be more options available for the buy-side in terms of execution venues and liquidity providers. The only way to access such a fragmented marketplace is electronically and through algorithms.”
Kedia says that the use of algorithms in FX is no longer limited to the hedge funds but traditional fund managers are now seeing the direct impact of recent FX volatility on their funds and are also turning to algorithms for smarter order routing. “It is no longer the case that they will do their FX hedge, or exposure, though a single trader at the end of the trading day.”
While FlexTrade provides a suite of algorithms, the platform allows users to customise or write their own proprietary algorithms. However, already there is a certain class of active and aggressive liquidity-seeking algorithms that are quite standardised.
Kedia adds: “Given FlexTrade’s strong background in equities it means that we have always dealt with latency, and take it very seriously, to the extent that our customers now say that the speed of light does matter.”
Co-location is where the banks and the hedge funds put their servers in the exchanges or co-location centres built by the network providers to minimise latency in accessing centralised markets. But as Kedia points out, it is only possible to co-locate in one location, which is not possible for the FX market, so co-location is really not an issue.
As he explains: “All you can do is find the best and the fastest network that can access all the trading venues. But beyond the network latency, there is still a bigger problem with the latency of the broker systems. In the equities world, the exchanges invest very heavily in improving and reducing the latency in their own systems because they see that the number of customers they have is directly related to latency with which they can process incoming orders.”
However, FX is equally competitive and Kedia says that the liquidity providers want to ensure that they are not disadvantaged by slower systems on their side, so latency is constantly being minimised and the use of algorithms is forcing greater efficiency. Additionally, the newer and fresher technology being used by the buy-side is also helping to reduce latency.
The key is using algorithms for smart order routing and Kedia says that algorithms should not only be able to find the best price, but be smart enough to know which destination has the lowest latency, to minimise the opportunity costs, and consider factors like the fill-rate and the hit-ratio when different venues are showing the same price.
Jim Feingold, global head of sales and marketing and co-head of product strategy group at Portware, says that banks are now less concerned about latency arbitrage, which arose when FX executable streaming prices became available a couple of years ago. They are also are getting better an employing preventive tools.
However, Feingold adds that because of the disparate execution venues for FX, the buy-side are looking for products that can not only aggregate the feeds but create a single display, similar to equities, which shows a top of book price, plus a synthetic depth of book.
He says: “The focus is currently on the aggregation of feeds from the best liquidity sources, and then because of the natural price scaling from best to worse price, and size offered at that price, they are able to create strategies and algorithms that take advantage of the best prices with most liquidity behind it.”
He believes the trend now is for banks to provide individual feeds to their clients and have much better control over their feeds and client relationships. “The banks pride themselves at being not only the best price in certain currencies, but also in providing liquidity in those currencies so a bank is confidant to be on an aggregated platform because they believe their prices will rise to the top.”
Feingold adds that by using algorithms to sweep the book for banks supporting size and best price, the FX market is starting to use equity-type algorithms to help them with the same strategies as the equities market, such as lessening the market impact, conceal the actual size of the order, and to execute size over a certain time period. “Just as the equities market use volume weighted average price (VWAP), volume is not reported in the FX market, the FX market has adopted a time weighted average price (TVAP) to time-slice FX deals and sweeping the book of the aggregated sources,” he says.
Feingold adds: “I think this benefits both side of the market. Because the buy-side is getting more confident in executing size, the banks are getting more flow, and the buy-side feels more confident in dealing in size because they are comfortable at knowing the price they are going to execute at.”
According to Feingold, Portware’s clients are aggregating bank feeds to create not only the execution-style strategies but also creating alpha, profit-making, strategies. “Black-box trading of FX has always been there, but now the buy-side is using tools to create strategies to react faster to the price changes in the aggregated feed. It started with the quantative hedge funds looking for alpha, but these strategies are now working their way into mainstream, driven by the demand for best execution, and the dealers are now creating algorithmic trading for FX, as well as for equities.”
Moreover there is a growing demand to also tie-in FX exposure to foreign asset class, trading instantaneously, and this too is driving the use of algorithms in order to hedge FX exposure risk when doing global equity or derivatives trading. This, and the simple fact that many of those looking to use algorithms for FX have experience in equities, means that FX is set to follow the same route as FX where automated trading strategies will simply increase year on year.