How AI Improves Liquidity in Credit Markets05.28.2020
Building Better Prices — How AI is Improving Liquidity in Corporate Credit Markets
By Ted Husveth, Director, Credit Product Manager, Tradeweb
How many traders, desk analysts and quants does it take to price a corporate bond? If you were to answer that question even a few months ago, the number could be as high as a half-dozen. Parties on both sides of the trade would be tasked with checking whether the bond traded recently, analyzing current credit and business conditions, digging into individual bond attributes and taking the pulse of the marketplace to see if the other side of the trade agrees with the price. For a complex trade involving a large portfolio of corporate credits, the process could have taken days.
Today, a single trader can do all of that in seconds thanks to advances in machine learning technology which have made it possible to calculate reference pricing in seconds based on dynamic bond market data.
And that is a huge step forward for liquidity in the $9.2 trillion U.S. corporate bond market.
The Corporate Bond Liquidity Challenge
The U.S. corporate bond market is the largest in the world, consisting of over 100,000 individual securities. However, less than 5% of these bonds change hands on a daily basis. That’s largely a function of the diversity of bond market. Unlike equity markets, where a single company has a single stock, in the corporate bond markets, a single company can issue several different bonds with variations including tenors, call dates, coupon structures, and contract terms.
That combination of diversity and illiquidity injects a great deal of interpretive fact finding into the bond pricing process. Add the relatively recent phenomenon of a surge in negative yielding corporate debt and the current “flight to safety” driving massive inflows into bond market mutual funds and ETFs and the challenge becomes clear. Time- and resource-pressed credit desks cannot afford to send out a small army of analysts every time they trade a bond. If they want to delve deeper into the corporate bond universe, trading anything other than the most-commonly-traded corporate credits, they up-to-the-minute pricing.
Pretium Ex Machina
While tech-enabled solutions have been flirting with the idea of real time corporate bond pricing for many years – through the development of FINRA’s Trade Reporting and Compliance Engine (TRACE), and various tools to catalogue where bonds have previously traded – the ability to predict where a bond will trade in the future, or provide accurate benchmarks for bonds that have not traded recently has been elusive.
Advances in machine learning have changed all that. The same variables that make corporate bonds so difficult to price manually are the textbook use case for machine learning, which is designed to process and parse vast data sets to find commonalities, outliers and trends.
Tradeweb has seized on the potential for this technology with its new Automated Intelligent Pricing (Ai-Price) technology, which delivers advanced reference pricing for more than 19,000 corporate bonds, updated every 30 seconds, using event-driven algorithms to determine where the market is at the point of trade.
The Ai-Price model ingests data from TRACE, information from Tradeweb Direct, Tradeweb’s real-time proprietary US Treasury Composite and CDS and US swap spread composite data from TW SEF to get a comprehensive view of where fixed income and derivatives markets are currently trading. It then applies an algorithm to consistently update all estimates with the latest market information, adjusting levels based on relevant new data, taking into account confidence levels, covariance and term structure. It also evaluates the size of a trade, counterparties and potential bid/ask imbalances to arrive at a highly accurate, real-world price in a matter of seconds.
A Unique Vantage Point
While advances in machine learning technology have unlocked the computing power necessary to tackle the bond pricing challenge, the real lynchpin to making Ai-Price work in the real world is the market data generated on the Tradeweb platform. Specifically, the Tradeweb Treasury and CDS Composites provide price transparency and serve as the basis for both transaction cost analysis as well as execution levels for AiEX trades in markets where Tradeweb helps facilitate to an average daily trading volume of roughly $120bn*. That’s a level of visibility into the pulse of institutional fixed income and derivatives markets that no other firms can deliver.
That deep, nuanced understanding of the real-time trading environment, coupled with powerful technology, is the fuel that makes it possible for Tradeweb to produce a solution like Ai-Price.
At a time when trading desks are running leaner than ever and living under constant pressure to keep trading costs as low as possible, Ai-Price makes it possible for traders to spend less time pricing and more time trading.
Technology has enhanced capabilities of surveilling larger and more disparate data sets.
With Eugene Kanevsky, James Redbourn, and Joanna Wong, CLSA
AI/ML on the buy-side trading desk is a long-term program with short- and mid-term deliverables.
Aiden uses the computational power of deep reinforcement learning to improve trading results and insights.
BondDroid’s AI-generated prices are integrated directly into LTX’s pre-trade analytical tools.