Feature: Fast Mover Advantage Published October 25, 2007 The Trade News
With high frequency trading comes a need for speed. Now long-only firms are under pressure to increase 'search' speeds to access liquidity across disparate trading venues.

Richard McClure

With today's trading landscape characterised by strategies in which humans are increasingly removed from the equation, and decisions are typically taken by computers in time periods measured in milliseconds, speed is more and more essential to maintaining a competitive position in the marketplace. As Philip Enness, global solution manager for financial markets data management, IBM, puts it, paraphrasing a slogan from the 1992 US election campaign: "It's all about latency, stupid!"

As the decision-making process becomes faster, the time to execute those decisions needs to be minimised. Latency, which stems from delays occurring at multiple stages in the trading cycle, is one of the primary impediments to high-speed trading. "You look at the whole trade lifecycle - from market data coming out of the execution venue through to the trading engines back through the sellside and back in as a trade - it's something like 200 milliseconds for the whole round trip," says Nigel Woodward, director, financial services, Intel. "So just shaving off five milliseconds here and there is what people are after."

This need for speed, says Woodward, is putting huge pressures on technology providers, exchanges and the sell-side as they compete in an arms race to come up with low latency solutions. If you are not in the top percentiles of latency infrastructure, warns Woodward, then the orders will be routed elsewhere. "Latency is of paramount importance to us - that's the reason we are in business," confirms Alistair Brown, founder and CEO of New York-based Lime Brokerage, which provides advanced electronic, agency only execution. "We cater to those traders that have no appetite for delays in getting their orders to the market - the hedge funds, the stat-arb guys, anyone who is doing any high frequency trading. When the market moves fast, they can't have any slowdowns at all."

As evidence of this low latency arms race, it is estimated that last year, firms spent £ 1.5 billion on trading technology in the City of London alone. The massive spend shows no signs of slowing, as exchanges and the sell-side continue to invest in speed. "We see latency as one very competitive component in our positioning in Europe," explains Michael Krogmann, head of section, sales, cash market development, Deutsche Börse. "There are some clients that really don't bother about the issues of speed - they have other business models. But other customers invest a lot of money into their infrastructure where milliseconds mean a lot of money. Being in the order book first is real money for them."

Growing closer

In April, Deutsche Börse upgraded its electronic trading platform, Xetra, with higherperformance hardware and new software functionalities. This means the average data transfer time for a round trip is now on average 16 milliseconds for the fastest participants. The exchange also reports an increasing demand for co-location, or proximity services - in which firms move the systems running their algorithms as close to the exchanges as possible to provide a shorter wire time and so further reduce latency. "We already have customers who place their systems next to ours. That will increase, because there are more and more market participants who realise that technology and speed is really a driver for their business models," says Krogmann.

Indeed, the exchanges are finding that the demand for server space in physical proximity to the markets can be a profitable business. In the US, many broker-dealers and execution-services firms are paying premiums to place their servers inside the data centres of NASDAQ and the NYSE, with approximately 100 firms now colocating their servers with NASDAQ. "Physically locating your machine in the geography that you want to trade within means you can trade competitively at the same latencies as the local market participants," explains Mark Akass, chief technology officer, BT Radianz, which announced a deal in April to offer proximity solutions to Deutsche Bank's clients.

In March, the firm also launched Radianz Ultra Access (RUA), a high-speed service that provides transaction speeds of approximately one millisecond between exchanges and brokers in the New York area. "Clearly there is a market and competitive pressure to turn things around quicker," says Akass. "Some exchanges are still matching things in tens and hundreds of milliseconds, and some are now getting down into a few milliseconds, so there is a spread out there already."

Yet while latency is an increasingly critical issue for exchanges and sell-side firms, it appears to be far lower on the list of buy-side priorities. As Phillip Hylander, co-head, European equities, Goldman Sachs, phrased it at TradeTech in April: "Given the choice between 300 microseconds reduced latency and 50% greater liquidity, most buy-side traders would opt for the latter."

Enness of IBM admits he "struggles with the concept" that the buy-side is not interested in latency. "Everybody needs speed," he insists. After all, as more liquidity is traded on fully-automated execution venues, automated trading strategies will require faster connections to compete. By 2009, it is estimated that over 58% of the buy-side in Europe will be using algorithms, up from 41% at present, with traders' needs evolving from traditional discretionary trades to a more 'in control' approach through the use of algos and direct market access to achieve better, more effective execution.

The weakest link

To succeed in this landscape, buy-side traders will need better, faster, more accurate market data. With electronic trading causing a surge in market data volumes, and traders seeking connectivity to liquidity pools, combined with algorithms and exchange venues moving into multiple asset classes, some buy-side firms are investing in execution management systems (EMSs). One of the main reasons for deploying an EMS is to have faster access to execution venues, an area in which order management systems (OMSs) often get bogged down with message traffic.

"The purpose of the EMS is to handle the communications directly with the exchanges or your destinations. That means its sole purpose is to get the orders out there," says Saul Nadata, manager, software development, Aegis Software. "If the EMS is doing it slowly, it doesn't matter what business workflow you put into place, it will be the bottleneck. In almost all cases, the EMS becomes the bottleneck, either in high volume trades for the sell-side, or on buy-side arbitrage scenarios where incredibly low latency is important."

Earlier this year, Nadata authored a white paper that provides a benchmark for measuring EMS latency, which he defines as the time from the receipt of a price update or execution report to the receipt of confirmation from the exchange or ECN. Although Nadata found that "generally EMSs have decent low latency," Alasdair Moore, director of sales at Fixnetix, believes latency is not the most important priority for some EMS providers. "I wouldn't say EMS providers weren't focusing on latency, but at the moment they are trying to integrate themselves into OMSs," he says. As such, adds Moore, they are far more interested in proliferating their solutions and making sure those solutions can be readily deployed in a number of different instances. "This slightly goes against building something for performance. You are either building something for performance or you are trying to make it easy to deploy with any other software solutions," he says.

However, not everyone is queuing up to buy an EMS. A survey of traditional fund managers by consultancy Investit in March found them divided over whether they should make an EMS investment or upgrade or retain their existing OMS. Does a buy-side firm need an EMS? It all depends on the dealing profile of the buy-side firm, and whether it is a high-frequency equity trading house or not, suggests Clare Vincent-Silk, a consultant at Investit. "As long as the buyside has a half-decent OMS, I don't see that they need an EMS," she says. "What they are after is connectivity to the different venues. It's not like the sell-side, where speed is of great importance. It is not so important on the buy-side. Your trading strategy is more important - what you are actually trying to achieve, rather than how fast you can do it."

Why latency matters

Thus far, latency tends only to be a priority for high frequency traders, such as statarb firms, and prop desks that are actively trading on their own behalf, with the need to utilise tools and the latest technology to compete. "Low latency is needed where it supports the trading strategy," says Akass of BT Radianz. "If you are doing arbitrage trading, for instance, latency is absolutely paramount because you are looking at gaps between pricing on different venues and different markets. Quite often that gap is where somebody else has not addressed their latency problems so they are leaving themselves exposed in the market and people arbitrage trade against that."

For many traditional asset managers, microseconds don't matter. What is far more critical to them is cost-effective, anonymous and reliable trade execution. As much of the buy-side is currently not conducting what would be deemed high frequency trading, they are subsequently less interested in speed as a solution. "Low latency is of far more concern to the high frequency traders - it definitely wouldn't apply to all of the buy-side traders," agrees Jeff Wootton, vice president of product strategy at Aleri, which provides complex event processing. "There are plenty of buy-side firms, not just the traditional asset managers but even hedge funds, that do not deploy a high frequency trading strategy."

According to Lime Brokerage's Brown, traditional buy-side firms are more worried about information leakage and getting a good fill than speed of execution. "They don't really care if their order gets to the floor in one millisecond or two seconds, as long as it's not 20 seconds or something," he suggests. "Latency is not one of their highest priorities right now. Their main concerns are information leakage and the right pricing." This, however, is to miss the point that low latency can bring competi- tive advantages even to traditional asset managers. As Brown puts it: "If you are fast, you will generally be getting much better fills."

One reason for this lack of interest is that many buy-side firms believe they have passive, rather than active, trading strategies. An example of a passive strategy would be a VWAP order, says Moore of Fixnetix. "For a VWAP order, speed is not actually that important", he suggests. "The problem is if the VWAP order is pegging a price to the best bid. By their very nature, most of the VWAP orders try to buy on the bid and sell on the offer, so if the order is being pegged to the best bid and the trader reacts one second after an update, they will be further down the stack in terms of the order than if they had reacted in, say, five milliseconds."

Consequently, because the order is further down the stack with more volume ahead of it, explains Moore, if somebody else hits the bid in a small number of shares the order won't get a fill. "Whereas if their algorithm had sent the order down earlier they would have bought those shares because they would actually have been hit. So if you react quickly and your order is positioned where you want to be sooner rather than later you are less likely to have to pay the spread." Fixnetix, adds Moore, is currently undertaking some empirical analysis in this area, with a view to publishing a white paper later this year. "It's not something that many buy-side firms have really analysed in very much detail," he says.

Enness of IBM makes a similar point. "Even if you are a long-only fund, then, yes, you can sit and buy the market and be in at a certain rate, say 10 for example," he notes. "If the market goes up to 50, you have made your 40 points, but if you are executing in a far more efficient manner, and you are getting in at nine, that's an extra point in your favour on your investment. Over a broad volume of executions over any given year, that point improvement will be significant."

With implementation of MiFID on the horizon, Ali Pichvai, managing director, Quod Financial, a cross-asset OMS/EMS provider, believes this buy-side indifference will end, as fund managers are forced to appreciate the need for speed. Once liquidity is sufficiently fragmented post- MiFID, he argues, the most important task for an EMS is to achieve best execution through fragmented liquidity pools, increasing the challenge of working out where best to route business, and making network performance to the venues a key component. With fragmentation, says Pichvai, the first task of a trader is not simply executing at the best price, but finding it in the first place in milliseconds amongst multiple liquidity pools, and executing orders dynamically as the market moves. "For exchanges and the sell-side, latency is the mantra, but this is not yet the case with the buy-side," he adds. "That will have to change when you are trying to grab the liquidity across different venues."

With continuing fragmentation and the increasing reliance on smart order routing technologies and algorithms to access liquidity, the buyside will have to look continuously to increase 'search' speeds and reduce latency when endeavouring to access liquidity to maintain a competitive offering. "Buy-side firms have been increasingly getting direct feeds from exchanges and ECNs. Those that have utilised those feeds well should already be benefiting from reduced data latency," says Louisa Paul, senior equity dealer, Baring Asset Management. "If latency isn't a big issue for the buy-side and head dealers don't have a handle on it now, they inevitably will have to soon."