• Related Articles

• 

  Customising FX algorithms: fine tuning your order execution strategies

• 

  Harnessing more intelligent FX Algorithms for superior order management and strategy execution

• 

  The next wave of FX Algorithms taking order execution strategies to another level

• 

  Identifying the optimal combination of FX algorithms

Build versus buy: choices for deploying advanced Algorithmic FX trading infrastructures

By Nicholas Pratt

Sullied Shakespeare soliloquies aside, the buy-versus-build debate continues to resonate in the financial technology space. "It's the classic question when it comes to technology and banks," says Giles Nelson, chief technology strategist at Progress Software, a provider of complex event processing technology. In the past, the answer to the questions has often been predicated by the size of the institution in question. The large banks have used their own considerable IT teams to produce their own systems while the smaller institutions have opted to seek help from third party vendors. But this is changing, says Nelson. "The big banks have vast IT resources and many would like to use these resources as much as possible. So in terms of FX algorithms they may want to look after the business end (the actual algorithms) themselves. But we have an increasing number of big banks that use our systems."

It can often depend on the culture of the banks rather than their size, says Nelson. For example, the likes of Morgan Stanley and Goldman Sachs are well-known for wanting to build everything themselves but the benefits of opting for a third party vendor are becoming increasingly attractive. "You get access to research and development and technical expertise. And why would you want to spend lots of money reinventing the wheel? Having said that, most businesses want to be able to implant their own intellectual property on a trading platform and to put their own business logic into the system so this means that they no longer want to use a black box," says Nelson.

By opting to buy rather than build technology, firms are able to do things more quickly, says Nelson. "This option also enables firms to protect themselves from a member of their own IT team walking away with the secrets to the technology they have developed and various other 'legacy' problems that we have seen come up in the past." It also puts the onus on the technology vendor to meet each clients' demands rather than selling a mass-produced off-the-shelf product with little customisation available, says Nelson. "The biggest differentiator for many vendors is the connectivity they offer. Can they connect the client to the venues they want? It is also important that clients are able to build the algorithms and decide on the trading strategies themselves and that they are able to see what is going on. This helps them to build the algorithms more quickly. For example, you can have different dashboards for different users - for traders, for quant analysts and for risk managers."

In today's software market there are systems being set up based on service-oriented architecture (SOA) where everything is component-based and while this may be good in certain areas, it is not a suitable model for FX algos because there are too many overheads which creates too much latency, says Nelson. However, we are seeing a greater use of hosted solutions and the software as a service/cloud computing model that is becoming so prevalent in the capital markets and other industries.

Giles Nelson

"The biggest differentiator for many vendors is the connectivity they offer. Can they connect the client to the venues they want?"

"This model simply means that the technology or infrastructure that a bank uses is maintained somewhere outside the bank and by a third party. This means lower cost and more flexibility and I see no reason why FX trading won't be influenced by the cloud computing model in the same way that other parts of the capital markets have been. In some ways it already is - for example, the single bank portal that is accessed through a web browser is another example of cloud computing. I think this model can reach the point where even the algorithms themselves are maintained off-site because there are various ways that the intellectual property can be protected without everything needing to be held on-site."

Hosted Solutions

Hosted solutions have certainly matured in the last 10 years and, in most instances, are far more appealing to trading firms than the application service provider (ASP) or service bureau model that were previously marketed to them. Back in these days, such models were principally aimed at the small and medium type of institutions that did not have the budget or resources to build their own system or to install and maintain a vendors' system. The ASP model allowed them to effectively rent a system that could be maintained by the vendor. But such was the level of customisation that many clients demanded and such was the burden of ongoing maintenance that vendors found themselves overstretched and unable to make the economics of this new arrangement prove effective.

However, today's hosted solutions represent a better deal for both vendors and customers, argue many software vendors, so that clients get both a reduction in price and also an increase in quality. "You get a solution very quickly and one that's tailored to your business needs. And the technology is now able to match any volume. It is a solution that does not fit all sizes," says Philippe Buhannic, co-founder, chairman and chief executive of Trading Screen, the US-based provider of multi-asset electronic trading solutions. The market conditions have also changed in ways that further support the case for a hosted solution says Buhannic. "If a bank decides to build its own solution it will have higher costs and more salaries to pay at a time when there is a great deal of cost pressure."

Philippe Buhannic

"The FX market is a lot darker than the equities market and this makes it more difficult to build algorithms. You have to have prices from multiple sources covering different regions and lots of information about trading flows and market movements."

As with many technology developments, both hosted solutions and algo trading have reached greater maturity in the equities market but they are still in their infancy in the FX market. "A lot of people in the FX industry are only just discovering algos and the providers of hosted solutions are still catering for a lot of customisation. But there are only so many ways to skin a cat, so to speak, and I think this will become simpler as the market matures and more algo trading toolkits become available that can cover all of the customisation that a client may want," says Buhannic.

As the solutions themselves improve, FX traders will also find the economics more compelling and the idea of building their own algo trading solutions will become less appealing. Furthermore, traders will be able to try out the hosted solution approach in a modular fashion rather than suddenly entrusting all parts of their trading operations to a third party. "There are different options available," says Buhannic. "You can have your network hosted, or your infrastructure or specific applications. Right now I don't see many fully hosted solutions being used in the FX market."

Hosted algorithms

Buhannic also feels it is unlikely that FX firms will be willing to use hosted algorithms quite yet. "The FX market is a lot darker than the equities market and this makes it more difficult to build algorithms. You have to have prices from multiple sources covering different regions and lots of information about trading flows and market movements. The intellectual property involved in the process is very important and this is what most trading firms are likely to want to keep in-house."

In the equities market, however, the algos are becoming more and more similar, which means that hosted solution providers are focusing increasingly on the implementation and the infrastructure around the algos in order to differentiate themselves. But the nature of the FX market, with its OTC relationships and multiple liquidity sources, means that there is still plenty of room for innovation when it comes to developing execution algorithms, says Buhannic.

Buhannic is less enthusiastic about the role of new technology such as CEP and .Net. "I think they are nice-to-have tools but there are lots of buzzwords about when it comes to technology. I don't think the whole process is that complex and perhaps vendors can sell their products for ten times the price by putting three letters in front of the system name. Technology can help but it is not the most important factor. It is no good if nobody wants to use the algorithm - that is the key objective, to make the best mousetrap. The advantage in today's market is that the same algorithms can look very different for different clients. Everyone uses them differently and has different architecture and infrastructure and supporting technology."

Service-oriented architecture is becoming a much-used term in the software world but Buhannic says that it is not a model that can really be applied to algorithmic trading because of its component-based properties. "We are using SOA in some areas such as integration and communication but it tends to be quite slow because of all the different parts and in algorithmic trading, the last thing you want to be is slow. We are using it for the stuff that is not time-critical - for pre and post-trade processes but not the trading itself."

David Hastings

"A lot of our clients do not have the expertise of integrating with other technology. It is a very complex area and it is continually changing, so for a firm looking to do this themselves, the platform can become very cumbersome or expensive to maintain"

Properties to look for in technology providers

So, for FX traders looking to employ the services of an algorithmic trading technology provider, what are the key properties to look for? "You have to look at the reliability of the provider," says Buhannic. "Do they have a reputation for service? If you pick the wrong partner, it can be disastrous. I have seen a lot of providers that take a long time to implement their systems, leaving firms to pay a lot of consulting fees. You have to make sure you know exactly what the system is doing. You don't want to undergo a massive process that in the end produces very little. The smaller providers are usually better at this but you cannot be sure that they'll always be around. People don't like to admit their mistakes so it can be difficult to ascertain what vendors are to be avoided. Ultimately you have to do your own research."

If this research is carried out correctly, firms can end up with significant advantages, says David Hastings, global head of FX sales at FlexTrade, a US-based developer of broker-neutral algorithmic trading platforms. "The biggest benefit of migrating to a hosted solution is that the whole thing will run by experts. The clients do not have to be experts in Linux or C++ coding and they can access a low latency system at the fraction of the price that would be involved in building or hosting their own platform. It is all about the benefit that comes from sharing the platform with other customers."

A further difficulty for clients looking to build or manage their own platform is that they will typically be using a range of technology from different providers and will have to make sure that it all works together in one harmonious trading environment - a task that cannot be underestimated. "A lot of our clients do not have the expertise of integrating with other technology. It is a very complex area and it is continually changing, so for a firm looking to do this themselves, the platform can become very cumbersome or expensive to maintain. By outsourcing this responsibility, they enable themselves to concentrate on the fine-tuning instead," says Hastings.

So if the economics of hosting their own platform does not tend to work out too favourably for trading firms, what makes the numbers work out for vendors? According to Hastings, the main advantage comes from having multiple customers. "The vendors operating in this space have to be cost-effective to be there in the first place and this is dependant on them having multiple customers."

Joey Horowitz

"We haven't offered hosted solutions in the past but it is something that we can now do because of the Thomson Reuters merger and the financial backing that we now have."

The advances in technology have also helped, says Hastings, particularly the development of the financial messaging protocol FIX. "We can use FIX to communicate with multiple users and to adapt the system to the specifications of each individual client. This allows us to have a more bespoke approach." Technology is also helping with the ease of implementation and the ongoing flexibility when it comes to customising the algorithms. "We can use a common interface based on Java or C++ and use remote servers to enable clients to access the algorithms when they want to through a simple GUI. This gives them more flexibility in terms of customisation and also puts all of the algorithms and the links to liquidity providers all in one place."

Of course, some firms can choose to retain the control of their algorithms by hosting them in-house and then look to a third party vendor for the links to the liquidity providers, says Hastings. But, he says, this still necessitates some technical expertise on their part. "Do they have that Linux coding capability, for example? Are they able to run the required disaster recovery capability such as the three separate lines of power supply? Do they have the right set-up for low latency? All of these things are costly."

So what kind of strategies lend themselves to the hosting model? "High frequency and high turnover trading is perhaps more applicable to the hosting model but the same distribution process is involved with any type of trading. It really comes down to what the IT resources are." And what should trading firms be looking for in a potential technology partner? "You have to be sure that they are able to meet your current requirements and also your future requirements," says Hastings. There are also more technical aspects involved in the selection of a technology partner. "How much is available straight out of the box? What interfaces does it support? How open are the algorithms and how customisable are they?"

Benefits of a hosted solution

According to Joey Horowitz, the chief technology officer at Aegisoft, a supplier of algorithmic trading software and direct market access, there are two primary benefits to a hosted solution and they can be applied to both large and small organisations. "If you are a large trading firm with a large number of users and several different offices in different regions, it can be a very complicated process to install a new trading system what with all the servers and all of the maintenance that is involved. A solution that is hosted by a third party makes this process much more manageable. It also allows clients to have much better control over the proximity of their trading system and servers, particularly if a firm is trading FX as part of a cross-asset strategy. So the benefits are both operational and strategic."

"A hosted solution is a choice that we will be able to offer our customers as a result of the investment that Thomson Reuters will be making following the recent acquisition of the assets of Aegisoft," added Horowitz.

Of course, as well as the fact that hosted solutions are becoming increasingly attractive, the task of building your own algorithmic trading system is becoming increasingly challenging and time-consuming. "Developing your own algorithmic trading system is difficult to do," says Horowitz. "Firms want to focus on the algorithm and not the plumbing but so much of it is plumbing. I know of lots of customers that think a simple spread algorithm is easy to come up with but it takes months to make that algorithm competitive and that's before even getting to the platform."

Harrell Smith

"In today's market there really is no competitive advantage gained from doing all of the development internally."

For example, says Horowitz, many banks offer a price band so algorithms can sweep the book but are only able to trade one band at a time, so how do you work out whether it's better to base on it on the book or the different bands? "A lot of this will depend on the level of technical expertise at each firm but you have to remember that the important point is the intellectual property - firms won't want to share that or have anyone else develop it for them. Having said that, I think every other aspect of the algorithmic trading process is potentially ripe for hosted solutions - the market data which is accurate, aggregated and up-to-date; the connectivity and communication; the logic used to co-ordinate all the different liquidity sources."

Challenges of building an in-house trading system

If a firm is engaged in high frequency trading, then it simply has to go with a third party platform, says Horowitz. "There is no way I'd start from scratch if I was an FX trading firm. To take this option, you have to be a firm believer that you can do a better job than those people that have been doing it for years and without having to spend loads of money. five or six years ago the only people building it themselves were the smaller firms going after one or two venues and picking up the cheapest technology parts. But I would never recommend taking this approach in today's market. There is so much continual investment involved today that I can't imagine anyone deciding to do this. I can't remember the last time a prospective client decided to develop their own."

Harrell Smith, head of product strategy at trading software supplier Portware, agrees that there are enormous challenges facing an FX trading firm that wants to build its own trading system. "First and foremost is the creation of a liquidity aggregator, which acts as a centralized FX trading portal - accepting and normalizing numerous data feeds, feeding that data into algorithmic engines, and receiving orders and routing them out into the marketplace. The process of aggregating and normalizing liquidity from various banks, dealers and ECNs is very complex and time consuming," says Smith. While most liquidity providers have adopted FIX as their standard communication protocol for FX, exact specifications vary from firm to firm and are constantly being updated, he says.

"Second, firms need access to real-time and historical FX data. Algorithms for listed products rely on the statistical analysis of reams of historical trade date, coupled with real-time analysis of current market conditions, executions, volumes, etc.

In the FX market, liquidity is fragmented among various non-public execution venues, so firms can't rely on the same kind of information that is available to them in the public markets. Third, firms need a powerful, event-based trading system that also provides a flexible algorithmic development environment. The latter allows firms to easily create and back test algorithmic strategies, and dynamically respond to changes in input variables and results."

Maintaining links

Another challenge facing firms that decide to take on the burden of building their own trading system is managing and maintaining the links to the all the various liquidity providers, says Smith. "All dealers' and ECNs' feeds are different. Some rely on FIX, while others rely on proprietary messaging technology. For those that have proprietary feeds, incoming message traffic and outgoing orders must be handled by customized API's and routing engines, respectively. Certifying, maintaining and aggregating all of these different quote feeds is a major undertaking for individual firms."

Of course, aggregation is only part of the battle, says Smith. "Once these feeds have been aggregated, firms must still integrate them with their algorithmic engines, and/or connect to a front end so that traders can see a consolidated view of all liquidity at various price levels. This requires significant customization of an existing front end, or the creation of an entirely new GUI. Neither option is particularly attractive, given the lack of flexibility in legacy trading architecture, the performance limitations inherent in these solutions, and the inefficiencies and costs associated with creating an entirely new FX trading interface.

"Today, however, firms can deploy a complete vendor trading solution that incorporates an FX aggregator, a fully customizable GUI and an algorithmic development environment. This approach offers firms the best of both worlds: extremely powerful out of the box functionality, the flexibility to customize any aspect of the system and the development tools necessary to create proprietary algorithmic strategies."

Building a high performance FX algo trading solution from scratch is an enormous undertaking, says Smith. And given the quality and reliability of vendor solutions on the market today, firms who would have previously chosen to build in-house, including some of the largest and most sophisticated quant funds, are going with advanced vendor solutions instead. "In today's market there really is no competitive advantage gained from doing all of the development internally."

Nevertheless Smith agrees that firms dedicate sufficient time and effort to the selection of their technology partners. "It is very important that firms select vendors based on their long term commitment to supporting clients. In terms of general system support and maintenance, this goes without saying. Just as important, however, is whether a vendor can be relied on to provide long term support for more advanced development and customization projects. Anything less is really considered outside the norm for established technology providers today.

"In addition to core performance metrics, technological sophistication and functional capabilities, firms need to carefully consider less obvious factors. Among these are a vendor's reputation in the industry, platform reliability, long-term value, and the extent to which a technology provider will be a true partner. The last issue is particularly important for trading firms who want to focus their energies on their core competency - developing profitable algorithmic strategies - as opposed to running self-contained IT shops, which is something that nobody is interested in doing."