The situation is so grave that today complex pricing engines are being placed into production at Tier-1 investment banks without the ability to test the pricing algorithms on anything but current market data, which quite likely will not cover all the scenarios to which the software is designed to respond.

What is needed is a new generation of market data test tools that allow developers to contrive particular scenarios within a sequence of market data updates. For example, these scenarios could be a shift in a yield curve, an arbitrage opportunity between a basket of stocks or a change in a seed price off which other prices are matrix priced.

Without sophisticated tools to contrive these scenarios and apply them repeatedly in a precise manner, developers are left to test their applications against live data which most likely will not contain the scenarios of interest. In addition, latency between input and output within these market data consuming applications has become critical. Without a known series of updates applied at precise market data rates, it is very difficult to understand what effect on latency a change in the code is producing.

In the general application development world, automated testing is considered a standard part of a modern software development methodology. Owing to the difficulty of producing repeatable market data streams, this rigor of automated testing has not been applied within the market data development world. Quite recently, tools to address this problem have begun to emerge that allow precise market data scenarios to be created and applied in a repeatable manner. These tools also allow for the market data rate to be varied, so both the content of the data as well as the update rate can be varied and applied in a repeatable manner. The tools create initial scenarios by recording a baseline market data stream and then permit programmable manipulation of the data to create arbitrary streams. These streams can then be applied in an automated manner at variable rates, while also including error conditions or system issues to be tested against.

Without these modern tools, searching for subtle errors in market data logic is not unlike a message in a bottle bobbing on the open sea, hoping to encounter a wave that pushes it in the right direction. With the profitability of the desk these days closely tied to electronic pricing and trading algorithms, being able to evolve this logic and test it efficiently is key to remaining competitive in this new age.

Howard Pein is the CEO of CodeStreet. To learn more about the above topic visit the CodeStreet web site at www.codestreet.com.