AFTAs 2018: Best Artificial Intelligence (AI) Technology Initiative—MarketAxess

With all the innovation currently taking place in the artificial intelligence (AI) field, this is one of the most keenly contested of all the categories on offer in the AFTAs. In 2017 the winner was FIS for its Control Center platform. The 2018 winner is New York-based MarketAxess for its Composite Plus (CP+) offering, a proprietary algorithmic pricing engine for corporate bonds. 

It was three years ago that MarketAxess first turned to machine learning. According to David Krein, global head of research at MarketAxess, machine learning has traditionally been more common in equities as opposed to fixed-income markets. “We sought to apply it to corporate bond market data, which is notoriously sparse,” Krein explains. “Market data can come from various sources. There is a lot of stale information, there is a lot of incomplete information, and a lot of it is not centralized. We went through a major data exercise to consolidate, condition and warehouse a lot of this data across the entire firm. Then we used machine learning to digest and make sense of it.” 

CP+, which can be used to support best execution and price discovery functions, was launched in May 2017.  The engine generates prices for more than 24,000 bonds globally from public TRACE data, proprietary MarketAxess trading platform activity, and Trax. Prices are generated up to every 15 seconds. 

According to Krein, the idea of having a pre-trade price to display took traders some time to get used to. “What we looked to do was something that hadn’t really been done before, which was to be able to use this price as a pre-trade reference for the corporate bond market, meaning that it should provide the greatest transparency by being closest in level—price, spread or yield—to the next trade that occurs in that bond,” he says. “Competing commercial products, certainly those that have access to more limited datasets and are not AI-driven, are not deployed as useful pre-trade references.”

MarketAxess continues to refine the machine-learning algorithm on a quarterly basis, incorporating new data points to improve accuracy and reduce errors. “There is a real inflection point in the corporate bond market right now with respect to data,” Krein explains. “There is finally enough information for us to build products like this and have them be meaningful. We feel that we started early in the cycle, made the investments in people and infrastructure, and because of that we are realizing the fruit of that now when the market needs it most.”