The Gateway to Algorithmic and Automated Trading

eFX: Speed or Bleed. Globally Fragmented Liquidity and Ultra Low Latency Forex Trading.

Bob Giffords

Forex traders knew all about fragmented liquidity long before the equity trading community had to grapple with the fragmentation problem. But the move first to electronic trading, followed by the arrival of high frequency FX traders has had its challenges. And opportunity springs from challenge. Bob Giffords speaks to the key players to find out how the FX trading community has adapted to an ultra low latency trading environment.

As traders go global, the foreign exchange (FX) markets move to centre stage, and speed is essential. "Algorithmic trading has changed FX as it did equities or futures," says Dave Rutter, chief executive of ICAP Electronic Broking. "Volumes are up and notional size, down. Traders also tend to feather their book with a range of prices in smaller amounts, so instead of gapping after major news announcements, we now see prices converge in a few seconds at the new level rather than taking many minutes."

"The market now reacts very fast to macroeconomic data releases," says Simon Jones, head of FX eTrading at Citi. "Twenty years ago over non-farm payrolls FX spreads would widen and it could be many seconds as prices became established - now it is sub-seconds before new equilibriums are discovered."

"Anticipatory models in the algorithms have also tamped down intraday volatility, increasing efficiency," adds Rutter. "The number of price levels traded during significant volatility is a massive difference to 2003-4 for example."

"Growth is driven by many factors: prime brokerage, technology, the growth of hedge funds and the retail phenomenon, especially in Asia," says Chip Lowry, COO of Currenex. "CLS plays a large part in driving a currency's trading volume. When Israel joined CLS, for instance, trading volumes tripled. We expect the same to happen for Chile and Turkey. The Turkish-Yen cross trade is already a popular retail play."

"The Eurodollar product is so commoditised now, traders are having to look at other baskets like emerging market currencies in Singapore, Hong Kong, and India," says David Hastings, global head of FX at FlexTrade, the specialist in broker-neutral trading platforms. "Indeed investors are starting to treat FX as an asset class rather than as an end of day funding activity." According to Hastings that means either managing a portfolio of currencies with a clear performance benchmark or using a dynamic, intraday hedging of FX as part of a benchmarked equities or futures portfolio. "These cross asset plays have grown rapidly," he adds.

"The traditional banking sector now accounts for less than half the flow that goes through the primary interbank venues," observes Jones at Citi.

"A few years ago the banks were typically price-makers, while the hedge funds were typically the price-takers," says Jim Kwiatkowski, FXall's head of sales for the Americas. "Now it's much more mixed and spreads have tightened. There's more liquidity across a broader range of currency pairs."

Jim Kwiatkowski

Jim Kwiatkowski, FXall

"There's more liquidity across a broader range of currency pairs."

"FX is not as transparent as equities," argues Nicholas Pratt, Associate Director with Lab49. "Some of the larger players still only distribute the top of book, not full depth, and throttle transactions. Meanwhile newer venues provide more information and don't restrict flow, so algorithmic trading is starting to grow. Once traders see liquidity move they will follow."

Pratt goes on to describe how the banks are trying to aggregate flow and push their own single dealer platforms. "They're also rolling out global risk systems to give a real time snapshot of their global positions and P&L across all desks," says Pratt. "As each market closes, its positions are marked, but their exposures will still be tracked." He also sees the whole process as self-feeding: volumes rise and spreads fall, which attracts more algorithmic flow. "Equity data feeds are now so fast, traders are using hardware ticker plant with FPGAs to keep up," says Pratt. "FX will be next."

A changing landscape

How have participant behaviours changed in this new global FX world?

"The biggest change in the market has been the shift of real money managers and corporates to more active trading," says Lowry at Currenex. "While they maintain their traditional bank counterparties, they are diversifying their liquidity pools." He notes how FX prime brokerage broke the link between trading and credit, and believes that central clearing will similarly break the link between settlement and credit.

"The growth of the retail trader has had a big impact," says Jones at Citi, "the presence of the 'Japanese housewife trader' in the market being well documented. Retail aggregators have sprung up everywhere and add a huge amount of liquidity to the market. This area will continue to grow as people look for alternative alpha in liquid products - the challenge will be for the reputable outfits to lead the way in business practices to ensure a few bad apples don't spoil the barrel."

"We see the retail aggregators using a combination of ECNs and streaming bank prices, which we can facilitate," says Kwiatkowski at FXall. "They typically trade in smaller size, but have traditionally used much more leverage, even 100:1. Regulators are bringing that down to 50:1 for US traders, so we'll wait and see how that plays out."

"Because there are few products and many participants, liquidity is very deep," says ICAP's Rutter, "which traders find difficult to push around. People can trade hundreds of millions of dollars without price impact. Still with high frequency traders constantly sniffing around thinly populated ECNs, fragmentation could bring new risks."

David Rutter

David Rutter, ICAP

"….with high frequency traders constantly sniffing around thinly populated ECNs, fragmentation could bring new risks"

"In the past all the banks used to make two-way prices for FX, while now only the larger players will do so," says Hastings. "Everyone else uses ECNs. However, now aggregators like FlexTrade are starting to become important and all the market participants need to engage with the aggregators to ensure they get fair exposure to the flow."

"Some traditional asset managers are now establishing FX prime broker agreements for a high frequency team to diversify their strategies," says Kwiatkowski at FXall. He notes that with the new Dodd Frank Act, prop traders are also beginning to leave the banks to set up new prop businesses on their own. So more change is on the way.

"On the sell side the banks are deploying increasing sophistication in their FX trading," says Hastings at FlexTrade, "by internalizing the flow whenever possible and distributing what they can't net across the available multilateral trading platforms."

"Some banks may try to internalize our flow by quoting just inside our spreads," says Rutter at ICAP. "Without post trade transparency that could erode spreads. Here the regulators are looking in particular at the single dealer portals."

"The banks may be internalizing their flow, but our volumes and transaction volumes continue to grow," observes Lowry at Currenex.

Flow may be fragmenting, but Rutter notes, "When the yen jumped 3.3% recently, we did seven to ten times our normal volumes, as liquidity flooded back from dark pools to our microsecond lit prices. People knew the risks and sought safety."

Tooling up the RoboTraders

How has FX algorithmic trading changed the trading platforms themselves?

"Currenex has a number of FX businesses," says Lowry. "We have our request for streaming quote (RFQ) model, popular with corporates and institutional investors who want to define the parameters for banks to quote, and our order-driven Executable Streaming Price model for active spot trading." Lowry describes how their ESP functionality allows traders to aggregate direct bank liquidity and the Currenex FXtrades anonymous liquidity pool. "FXtrades is aimed at the high frequency trader or bank prop desks with a central counterparty providing anonymity and settlement," says Lowry. "We offer various order types with this Currenex ECN, including icebergs and execution algorithms, covering any spot currency pair that clears through CLS. We also offer both classic GUIs and software APIs."

He also notes that Currenex provides its technology to some 30 banks and FX aggregators on a white label basis. "They source their own liquidity or may be part of the Currenex anonymous liquidity pool," says Lowry. "It is a fully hosted solution, so the deployment of these platforms is almost immediate. A number of retail aggregators trade on margin, rather than use credit. They're turn-key."

Kwiatkowski recalls when FXall launched its anonymous, order-driven ECN in 2007 using a prime broker model which he says appealed to the asset management and hedge fund community. "Our low latency architecture significantly leveled the playing field for buy and sell-side traders," says Kwiatkowski. Last year FXall acquired the LavaFX ECN from Citi. "Our goal has been to deliver the best of both platforms in terms of liquidity and technology to the combined FXall ECN and LavaFX ECN clients," says Kwiatkowski. "We will complete the migration of all clients to FXall's Order Book platform by the end of January."

He describes how high frequency traders want low latency and high volume of activity. Meanwhile, active traders looking to earn alpha want displayed and dark order liquidity pools, with smart orders that will peg to the best bid or offer, perhaps with an offset, and have the ability to match against a range of prices. "Traditional asset managers also find our TWAP and other algorithmic execution functionality useful," says Kwiatkowski.

He believes FXall has a unique business model that combines their ECN with a multi-dealer aggregation model and their market-leading RFQ (request for quote) platform. "Our Aggregator includes Bank Stream for disclosed, relationship based streaming prices provided by 25 banks, so traders can mix anonymous ECN liquidity with streaming bank liquidity. Because of the extremely low latency in our matching engine, some banks use our technology on a white label basis for their own internal crossing engines."

ICAP has adapted in a different way to cope with their huge volumes and global demand. "We have a unique single platform with three matching engines in New York, London and Tokyo," says Rutter. "Prices are published to all markets 24 hours per day. We have a 250 millisecond minimum order time against a maximum 140 millisecond price round trip across all platforms, so traders can't flash a price in New York without being at risk of transacting against an order in Asia etc. We also throttle order submission to keep orderly markets."

Ever faster, ever smarter

With all this innovation, it is not surprising that algorithmic demand has grown. "FX is very attractive to high-frequency traders because of the liquidity, multiple price sources, low transaction costs, good levels of leverage, and, most importantly, the market is far from saturated, particularly in places like Brazil or Asia, says Ram Rao, director and head of sales and business development at Trading Cross Connects (TXC), an incubator company for high frequency traders. "We know many equity firms planning to try out their models in FX."

"In the last couple of years," continues Rao , "as high-frequency volumes have grown and technology has improved, the banks have dramatically changed their attitude and are now actively courting algo traders with execution and prime brokerage services. We also expect more ECNs to offer rebates to market makers at the expense of price takers as we saw with equities. So competition should intensify to improve liquidity and tighten spreads."

"For high frequency strategies traders need to be globally situated and in hosted environments close to the markets," says Hastings at FlexTrade, "to take advantage of the tightest and deepest sources of liquidity."

"Most hedge funds are utilizing extra-net or using fast links," observes Lowry. Currenex uses its own Scratch protocol for market data and does not throttle their data. "Opportunities for arbing continues to exist," says Lowry, "but I think we're beginning to see diminishing returns because of the equal distribution of this leveling technology in FX. Traders will have to look elsewhere for advantage."

Besides fast networks compute power is often key to successful algorithmic trading. "There's a lot of mathematics but so far it's been straightforward," says Pratt at Lab49. "Monte Carlo simulations can be easily scaled up using a larger grid, but if you can do the calculations faster, there's always more potential for alpha. Moreover, when volumes spike, it's easy to get out of position as the latest algos look at prices in context. Yet no one seems to have had success with GPU chips, for example." If they have, then Pratt thinks they're keeping it secret. "And it's secret for a reason," he adds. "I'm not aware of a good commercial hardware product that's using GPUs. Based on my knowledge, this has been an avenue that has been explored by commercial vendors, hedge funds and banks, but FPGAs combined with general purpose CPUs are the current best of breed. So the arms race goes on and the focus now is on complex event processing or CEP, which is ideally suited to these kinds of problems."

The other big technology focus has been on risk management. "We've invested heavily in aggregation and risk technology," says Jones at Citi, "and the battle to ensure we are keeping pace with our clients and competition is never ending."

"In the last few years, top tier banks have been gearing up their front and back office to provide multi-currency, multi-entity, cross-asset, real-time monitoring of positions and counterparty credit limits," says Chris Davis, co-founder of TwoFour, an FX cash and risk management solutions provider. "They check off confirmations, posted margin, and net settlements, as they happen, and calculate a continuous P&L, balance sheet and cash ladders, with alerts as required. This is really now essential for today's high-speed traders and it's what we do."

If you are doing a thousand trades a second across multiple venues, Davis believes you have to know where you are long or short as it might influence order routing decisions, trigger back-to-back trades or intraday funding calls. "Broker-dealers are even using our systems for pre-trade due diligence," says Davis, "linked to their real-time pricing engines, OMS and trading portals, and then applying trading assumptions and what-if scenarios."

"Being able to see a client's total FX positions across DMA, broker algos and dealer blotters on all desks in the context of your own FX positions, while keeping pace with their high-frequency trading styles, really improves critical decision-making to allow them to continue trading or cut them off, for example," concludes Davis.

Some things never change

"E-commerce doesn't seem to have damaged the asset manager's relationship with the sell side," says Hastings at FlexTrade. "Indeed, sales traders are in demand as never before as the markets become more fragmented and complex. Buy side best execution and counterparty risk are still the two key factors, but both need to be analysed within the wider P&L context. Traders may now want to preference the banks based on the business and information they've delivered, for example, by offering depth even if it's a shade below top of book. So an FX EMS needs to provide the necessary real-time TCA analytics to achieve this."

"Compared to the traditional relationship model, the anonymous, transactional approach of some high frequency clients is much more challenging for the banks as market makers," says Jones at Citi, "and forces us to examine our approach to risk management."

Yet despite all the revolutionary change and growth in FX of the last couple of years, still more may lie ahead. "Non-delivered forwards and swaps are derivatives, so the new OTC clearing rules might apply," says Rutter at ICAP. "If interest rate swaps were cleared as well, that could open up trading to the masses and turn-over would soar."

"Whatever happens," concludes Rutter, "traders will continue to deepen their analysis of book shapes, news events and economic releases with ever more sophisticated models. There's no turning back."