The Gateway to Algorithmic and Automated Trading

Building and maintaining an algorithmic trading tool

Published in Automated Trader Magazine Issue 04 January 2007

As the use of algorithmic trading tools has become more widespread, clients’ expectations have similarly increased. Owain Self of UBS highlights the important considerations when developing and supporting a credible algorithmic trading tool and what it takes to maintain that credibility in a constantly changing market.

Owain Self

Owain Self, Executive Director, European Algorithmic Trading, UBS

In the fledgling days of algorithmic trading it was possible to buy or to build a system on a very affordable basis. But, as usually happens, it was not long before the realisation hit that you would get what you had paid for. Clients' expectations have since changed.

The initial frenzy among broker-dealers to have any kind of algorithmic capability within their portfolio of trading tools, regardless of the robustness or performance of the system, has given way to an increased level of circumspection. And those providers that were offering affordable but less effective systems have been found out.

In today's markets, clients are looking for performance, flexibility and reliability - attributes which require an investment running into the tens of millions and a worldwide team of people that exceeds 100. This realisation has limited the market to a select and credible group of five to six major, global broker-dealers who are willing to make this investment in technology and expertise. But what does it take to reach that elite group of providers, how should that investment be spent and what work is needed, in a trading discipline where performance and capability must constantly be improving, to maintain a position at the top of the algorithmic trading table?


The first investment must be in assembling a team of highly qualified experts. There are three areas to draw from - the traders, the quantative analysts and the technology developers - and it is essential to create a balance between these three groups. It is unrealistic to expect to be able to recruit an all-rounder who is the best-in-class for all three groups and it is also unwise to invest too strongly in bringing in a technology capability without having a similar pool of talent in the trading and quantitative areas.

There are clear responsibilities within these three groups - the traders will be more experienced in the end-users' behaviour and what they would like to see in terms of functionality. However, this does not mean that the development process should be reduced to the traders producing a list of functions and features and then expecting delivery of their dream some days later.

The quantitative analysts are becoming increasingly more important in the development of algorithms, as the models and the understanding of risk takes on more sophistication but a successful system will not be one that runs the most mathematically sophisticated modelling process but is nigh on impossible to be used by the average trader. Everybody has to be involved in the decision-making process.

At UBS, and as a global investment bank, we are able to draw from our considerable skill base in-house and select people that show the right blend of expertise that can be brought into a team that will develop a process where input comes from all three sides - the technology, the traders and the quantitative analysts.


Development of the original system is only the beginning of the project and at UBS it is difficult to pinpoint that singular moment of genesis. It has been a continual progression from the late 1990s onwards to the point where we are now on our third generation of algorithms. Technology obviously plays an essential part in a process so dominated by performance and a considerable investment has to be made. The latest generation of technology is essential and legacy systems are to be avoided - particularly when constant development and improvement plays such an integral role in the product's success.

Whereas the original algorithms in those first generation products in the late 1990s were more rudimentary and could theoretically have been developed by a trader, the ongoing sophistication in trading and modelling means that there is now far more input from quantitative analysts and a necessity to find the most suitable technology as opposed to a standard software package or an off-the-shelf piece of hardware. Each new generation of algorithm trading tools, however, is not built from scratch. It is about finding the limitations of existing systems and then making the necessary improvements so it is often a case of moving sideways in order to go forwards.


Changes in the marketplace also necessitate constant redevelopment and for this client feedback plays a vital role in development efforts. However, to ensure that the maximum benefit is derived from the feedback, it is important to look beyond simply acting on verbatim customer comments without setting them in any kind of context.

The level of expectation and of education must be considered alongside any clients' comments. For example, a client may not appreciate the way a process is performed because it is simply not the way they are used to working - which does not necessarily mean that the process is wrong or in any way inferior. Working constantly with the client so that there is a true partnership rather than a one dimensional client/vendor relationship helps to determine what feedback can be useful for future development as opposed to a 'what the client wants the client gets' dynamic.

It is also important to rely not solely on your customers for their feedback. By sitting waiting for this feedback to come in, the lead time involved would mean that none of these changes would be implemented on time. We would also all be developing the same product because every provider would know what these clients' demands were. In a fast moving environment such as algorithmic trading it is vital to stay one step ahead not only of competing providers but also in trying to anticipate your own clients' expectations and feedback.

This can be generated internally from the banks' own internal trading teams and from the development team itself.

Other considerations

While the size of investment and the level of expertise are strong determining factors in the success of an algorithmic trading product, there are other factors to consider. The ability to fulfil all of a client's trading ambitions is as important as is the level of customer service in that there is someone on the end of a phone for clients at all times.

An algorithmic trading tool is not a stand-alone system so it has to be able to integrate with other systems within the trading departments and also the back and middle-office processes. Global reach is also important for those operating in multiple jurisdictions. During the development phase we like to work with regional experts when looking at products for individual markets because they understand those markets better - as opposed to developing one monolithic product that you then try to adjust for each market.

It also important to understand and appreciate the regulatory developments that are taking place in each region; that it is possible to build any of these changes, as well as general changes in trading strategy, into an algorithmic tool. For example, with MiFID taking effect in Europe next year and similar developments happening in parts of the US market, such as Reg NMS, traders will be looking to access multiple venues. Speed is important to a degree, although not as important as say direct market access in terms of execution. But the fact that UBS is an exchange member in most markets is a big advantage over some brokers' offering algorithmic tools.

The final advantage is the anonymity which clients have using the trading system and the reduction of information leakage. It is important that the technology and controls are there to ensure this level of security while still maximising crossing opportunities within the investment bank. Clients' confidence in their broker increases when that broker is more open how internal trading desks use the algorithmic trading systems. Whether the user is an internal or external client, their access should be the same.

It is a constantly evolving process and once an algorithmic trading system has been built, one can never sit back and rest on contented laurels. The development is constant. It is day by day, minute by minute and millisecond by millisecond. That vested effort involved in tracking market and regulatory development and clients' requirements will never slow down.

© UBS 2006. All rights reserved.

For further information please contact: email: