The single most important ingredient for success in securities and derivatives underwriting, trading and investing is complete price data, which in terms of granularity, is called tick data. Perfection means being able to see every trade as it occurs, knowing whether that trade is an offer taken or a bid hit, and knowing how much volume was done with each trade as soon as it occurs. It also means keeping this tick data in an easily-accessed historical data base.

BNP Paribas, now one of the largest commercial banks in the Eurozone and the sixth largest bank in the world (in assets), knows this of course, but in the global fixed income markets, this is no simple feat.

Many of the standard tools of technical analysis were first developed for use in commodities that are now called futures. All futures are traded on exchanges, which makes tick data collection an easy task.

With these data series and computers, one can apply even the esoteric tools of technical analysis like an Accumulated Swing Index, Aroon Oscillator, Blau Ergodic Mean Deviation Index, Parabolic SAR (stop and reverse), and Williams Percent R to name only a few, as well as simple, smoothed or weighted moving averages to make better trading and investing decisions.

BNP Paribas’ global fixed income division is comprehensive. “We cover all fixed income markets, which for us means all foreign exchange and interest-rate products, bonds and credit derivatives,” said Xavier Pujos, deputy chief operating officer in London, “but doesn’t include commodities.”

“We have aggressive plans to maintain our leadership position by applying sophisticated modeling and analysis to value fixed income products, uncover trading opportunities and widen our adoption of electronic fixed income trading,” Pujos said last December. Indeed, any securities dealer today must be able to do these things.

Thirty years ago, Harris Bank (now Bank of Montreal) was the first primary dealer to post its on-the-run, live US government bond bids and offers on Reuters screens throughout the world. Mel Swanborn, who ran the US government for Harris Bank, and who was criticized for it by some other dealers once told this reporter, “Like some fish, we bond dealers have to keep swimming in terms of evolution, or we will sink.”

With accurate tick data, one can also apply proprietary algorithms and even automated trading systems as well as the standard tools of technical analysis, something all bond dealers have been trying to do for years.

However, fixed-income and foreign exchange are usually traded in over-the-counter markets, that is, dealer-to-dealer rather than on exchanges. This means that capturing reliable tick data from perhaps more than 100 different sources as they occur, and collating them in exact chronological order is a formidable problem.

Searching for a Solution

Last December, BNP Paribas announced that it had selected the Reuters Tick Capture Engine (RTCE) to provide historical and real-time data for quantitative research and trading analytics on global fixed income markets.

“We first began to look into this by talking to Vhayu Technologies some 18 or so months ago,” Pujos recalled. Vhayu, headquartered in New York, offers a patented software platform with a calculation engine built on a Windows platform that can handle up to 400,000 updates of market ticks per second.

Pujos relates that BNP Paribas already had a solution with end-of-day foreign exchange data, that is, the open, high, low and close for each currency at the end of each business day. “From that, we evolved into trying to add some tick-by-tick data which was not an easy way of doing it,” he recalled. “You have to kind of use a shoehorn to do that, even though FX data are relatively clean and continuous.”

In June 2005, Microsoft Gold Certified Partner Vhayu and Reuters entered into a worldwide agreement in which Reuters distributes and supports Vhayu software that was optimized for Reuters data.

BNP Paribas’ research and trading teams saw quickly that they needed a somewhat different solution. When they learned of the new Vhayu relationship with Reuters they started looking into the Vhayu product. They began serious discussions in the middle of last year. “Signature of the contract to delivery took, I think, three months,” Pujos said. “We went live in November, 2006.”

Through a Microsoft Windows Server System, this Vhayu platform can scale efficiently through volumes of data, select relevant data and apply both standard and custom-made analytical tools to calculate solutions with response times of less than 20 milliseconds. The server allows multiple users to run Matlab, S-Plus and Excel programs on their individual desktops or workstations.

The newest version of Matlab, R2006b, was released last September. It is used to run analytical programs and has a graphical interface for solving optimization problems. This means minimizing or maximizing a trading algorithm by testing variables within a set range to find the best (optimal) solution. Users can build and/or test various strategies for one that can best solve a dealer’s or customer’s problem. Matlab can also verify that these solutions conform to specific standards.

0S-PLUS7 is a predictive analysis program that can handle very large sets of data and complex analytical business processes to present statisticians, analysts and researchers with solutions. It can be used to develop specific applications and permits non-statisticians and decision-makers to do their own analyses.

“You can also access data though Microsoft Excel,” said Donovan Ransome, director of channel marketing for Reuters in London, which provided the principal component for gathering prices, the RTCE.

The changeover from BNP Paribas’ old system meant migrating the data, the technical tools, and the proprietary algorithmic functions without interrupting daily business. In addition, Paribas could not stop collecting data. “Some of this is still going on because a few of the most complex add-ons still need to be coded properly and optimized,” Pujos explained.

Other Parts Added

The basic tool of this solution, the RTCE, was first launched in mid 2005. It can gather and store large amounts of streaming (live) and historical bond and FX data from multiple market sources including OTC data. It can also capture volume for each trade (tick) when it is available.

“The RTCE engine uses a client’s adapter, if it is available, to normalize inputs into protocols that it understands,” said Remy Grandville, Reuters marketing head for France and Benelux in Paris. “So either the client does the integration work, or the RTCE will do it for him.”

“RTCE is flexible enough so that trading strategies developed by a quant can be hosted in a second analytical engine with real-time data that updates this strategy each time that security’s price is updated,” Grandville explained. “The strategy can then be distributed throughout the Reuters platforms to the workstation of each trader.”

All of the tools described above interface readily with the RTCE, which gathers some of its data from the Reuters Market Data System (RMDS). This platform has access to more that 70 exchanges, brokers and data vendor sources and it also collects live prices from Paribas’ fixed-income division as it does its own trades.

The RMDS avoids “tick mirroring.” If the price comes from two sources, one of which is the buyer on one side of a specific trade and another the seller on the other side, the trade is not duplicated.

For back testing and other historical studies the Reuters DataScope keeps 10 years of tick history. This entire solution runs on Intel Xeon processor-based server platforms.

This solution does not store customers’ portfolios to facilitate bond or derivative swaps, nor does it process tickets. “It just gives access to tick data,” Pujos explained.

Avoiding Copyright Infringements

One constraint BNP Paribas had to consider was making sure that real-time data is accessed only on screens for which the bank pays copyright fees.

The final piece is the Reuters Open Data Access Control System (OpenDACS) that simplifies managing data infringement liabilities. One way that exchanges make money is to sell live data on which they hold a copyright. Data that is 20 to 30 minutes old usually reverts to the public domain like Rimbaud’s poetry or Shakespeare’s plays on which their copyrights expired long ago.

On younger data, exchanges charges are usually assessed per screen ranging from $50 to $70 per month for each exchange. Applications using copyrighted data are controlled by OpenDACS so that these data cannot be routed to unauthorized PCs or workstations.

The imposed constraints on the publication of live data is a problem that many in the securities industry would like to solve. How live data can be distributed to customers and regulators to prove that securities dealers have done the best executions of an order is one issue. A larger question is how market activity can truly be transparent immediately if live data is restricted.

“It is an important issue for equities,” Pujos added. “It is not as important for fixed-income yet, although it could be in the future.”

Meanwhile, BNP Paribas is replicating the solution it has crafted with Reuters in Europe for its office in New York, so that the entire fixed-income operation will be on the same system very soon. With it, BNP Paribas believes it can maintain and improve its position as a dealer in the little-known and complicated world of bonds, foreign exchange and their derivatives.

www.bnpparibas.com

www.reuters.com

www.vhayu.com

By Desmond MacRae