The Gateway to Algorithmic and Automated Trading

Optimising Market Data for Buy-Side Automated Trading

Published in Automated Trader Magazine Issue 09 Q2 2008

The next generation of buy-side trading models is driving new requirements for market data. Buy-side firms use market data in a variety of different ways and for different purposes, but how asks Valerie Bannert-Thurner, Managing Director of Skyler Technology Europe can they make the most of the market data available to drive their automated trading decisions?

As automated trading increases in complexity, the need for faster, more comprehensive, and differentiated market data becomes more pressing. By leveraging the growing array of market data sources available today, buy-side firms are faced with a new opportunity to innovate, and ultimately profit.

The traditional way of using market data for automated trading relied mostly on pricing and pre-aggregated information such as Level I tick or quote data. Real-time tick data was used to trigger orders and historical tick data for model generation and backtesting. But with automated, model-driven trading going mainstream, this approach offered little opportunity to generate and leverage a competitive advantage, as everyone had access to the same limited pools of data. It also failed to provide usable information about the overall market structure - something that is becoming increasingly important in today's fragmented environment. As a result, the automated trading models that leverage this data have tended to be simplistic and homogenous in their approach to using market data, limiting the alpha that they can generate.

However, the market data available today is much more comprehensive and diverse, ranging from full-depth liquidity information, over news information to transaction cost analysis (TCA) reports. This enables the automated trading engines that use it to be much more comprehensive and complex in their approach, and ultimately more profitable. But are they making the most of the data available to them?

More than just speed

As markets became more electronic, the resulting 'need for speed' had a major impact on market data and the infrastructures within which it is distributed. Markets became faster and speed more critical, giving rise to direct exchange feeds, co-location and the use of hosted data centres. In addition, a technology arms race for a faster market data infrastructure was spurred. Yet these developments were predominantly focused on Level I tick data, resulting in a one-sided view of the trading activity. This approach lacked differentiation, did not reflect a comprehensive view of the market activities and was agnostic with regards to execution quality delivered by the different order execution mechanisms. Competitive advantage could be derived simply from the speed of your market data, but not necessarily from the type of the data itself. The market data environment and the automated models that it powered became fast, yet remained relatively unsophisticated.

“… buy-side firms need to focus on comprehending overall trading activity by incorporating more data from different sources into the model, …”"… buy-side firms need to focus on comprehending overall trading activity by incorporating more data from different sources into the model, …"

Today, we are seeing that speed alone is not enough. Models are becoming increasingly sophisticated. With the need for speed almost pushed to the limits, differentiation and alpha in automated trading can only be achieved by developing smarter models than the competition. The reason why alpha continues to decline in many automated trading strategies is simply because emphasis with regards to market data has been placed almost purely on speed, regardless of the fact that the speed alone is not the key to alpha in today's markets. Tweaking the tick-data-based models is not going to be enough, buy-side firms need to focus on comprehending overall trading activity by incorporating more data from different sources into the model, such as combining changes in liquidity with tick developments.

In addition, buy-side firms are becoming increasingly 'execution aware'. Initially, order execution was a service provided by the broker. As hedge funds and other buy-side firms are becoming increasingly independent from a single broker and connecting to a multitude of investment firms, their choice of execution mechanisms has exploded. Hundreds of different execution algorithms can be used to execute an order and buy-side firms are becoming increasingly aware of and sensitive to the differences between the brokers and their algorithms. The choice of the right execution route becomes a critical one and is seen as a core competitive advantage. This decision of which execution algorithm to use is increasingly incorporated into the automated trading infrastructure, with models becoming increasingly aware of the overall market structure and the efficiency of the different execution channels within it.

Smarter automated trading models

The buy side needs trading models that are smarter and can differentiate themselves in two key ways: firstly, through the use of differentiated and more comprehensive market data; and secondly, through increasing 'awareness' of the execution mechanisms and market structures in which they operate in order to intelligently route orders to the counterparty that will execute them most efficiently.


1. Smarter trading models based on fast, differentiated and more comprehensive market data -
This trend results in two key changes to how underlying market data is being used. Firstly, we have moved from single to multiple view points. Buy-side firms are increasingly complementing their tick data with additional market data such as Level II order books (liquidity information) to gain a more comprehensive picture of overall market activity in order to trigger smarter trading decisions. This applies to historical as well as real-time data and means that the trading model considers more than just a single view in terms of price movements. It also considers multiple perspectives on the trading activity, e.g. 'Is it possible to relate changes in the liquidity to price movements?' These smarter trading models consider the entire activities of the market including developments of the ticks, as well as changes to liquidity during the decision-making process. Additional views or perspectives are added through the inclusion of more sources of market data. For example, recently news sources such as Dow Jones and Reuters have been added to the market data sources used by trading platforms.

Secondly, we have moved away from using only external public sources, to using sources from all sides incorporated into the overall trading process. Until now, market data had only been available from public market data vendors and the exchanges. These non-proprietary sources are complemented by data from brokers in the form of indication of interest and TCA reports, as well as exchanges, multilateral trading facilities or even dark pools in the form of execution or trade reports. This data is proprietary to the specific buy-side firm and can therefore constitute key competitive advantages over other firms. In addition, buy-side firms are applying increasingly sophisticated analytics over all their data sources and can then extract an actionable 360° view of market activities from a list of multiple sources. Therefore, market data is becoming increasingly comprehensive and a source of competitive advantage by integrating not only public but also proprietary sources into the overall views.

"To support these more sophisticated automated trading strategies the buy-side market data infrastructure needs to be altered."

The implications for market data infrastructure are far ranging. Not only do systems need to deliver in terms of low-latency data processing, they must also complement the tick analytics components with analytics of Level II order book data and execution information within a single platform.

2. Smarter trading models that can evaluate the execution options of all your different brokers -
Whereas brokers use smart order routing technologies to achieve the most efficient execution across different venues, buy-side firms are concerned with the execution efficiency of one broker-provided execution algorithm versus the other broker's algorithm. Buy-side firms are swamped with a choice of different execution algorithms with a multitude of different parameters. These algorithms vary in terms of their execution logic or their access to different pools of liquidity, including dark pool algorithms. Quite often the choice of algorithms between brokers A and B can be a difficult one. Even TCA reports cannot entirely answer the question of which algorithm to use, because comparison of the different TCA reports can be somewhat challenging. There is a need for buy-side trading models to facilitate the decision of which execution route to take - put simply, a buy-side order router that does not choose between the available venues, but instead, the brokers and their respective execution algorithms. These buy-side order routers will complement and support the traders in choosing the most efficient execution route for a given security under certain market conditions. From a market data perspective, this means the data will not simply inform about securities alone, it will also inform about the execution channels to trade these securities. TCA reports will have the data available to compare algorithmic performance. Real-time trade reports and IOIs point to the availability of partly non-displayed liquidity at certain execution venues accessible via specific execution algorithms. These proprietary information sources need to be combined into a single comprehensive picture to support buy-side firms' decisions as to which execution channel to use and therefore how to achieve a competitive advantage in automated trading using market data.

Market data is the enabler!

In short, automated trading models used to be fairly slow and dumb in terms of leveraging the market data available. Over time, they have become faster and more sophisticated. As the demand from the market increases and market data becomes ever-more comprehensive - especially in terms of the growing availability and use of multiple sources (from public information sources, brokers, internal analytics etc.) - buy-side firms are able to develop more efficient and complex models. But more importantly, by leveraging the growing array of market data sources available and developing their own unique pictures of the market, the buy-side's automated trading models are again becoming true means of differentiating themselves from their competitors. To support these more sophisticated automated trading strategies the buy-side market data infrastructure needs to be altered. The infrastructure needs to be modular and scalable and allow for the integration of multiple data sources and associated analytics capabilities, while providing a single interface to downstream trading engines. Its key functionality will go beyond connectivity, access and normalisation to smart aggregations of different data points into actionable views that inform overall trading activity and associated execution options.

The sophistication of managing market data and adding multiple sources of information into a single comprehensive view of market activity, combined with efficiency of execution, will make all the difference for buy-side firms in the future.