One of the last human points of resistance to the algorithmic take-over of professional trading can be found in the fixed-income trading desk. The deft relationship business is known for having participants with Ivy League pedigrees and a lacrosse playing background, a data point that a machine learning algorithm recently used when making hiring recommendations.
With an algorithmic transition, profit margins are being squeezed as human relationships are becoming less important. Benchmarking the transition are new hiring mandates at some of the largest banks, a recent Greenwich Associates report pointed out. Relationships and fundamental market understanding are less important, the report noted, and algorithmic experience and data science skills are now in demand.
In this emerging environment, what are best practices for an integration framework?
When combining fundamental investment thinking with quantitative thinking unique conflicts often arise. Fundamental, discretionary fund managers often creatively connect noncorrelated dots and sometimes challenge consensus thinking. Quantitative logic, on the other hand, is often driven by if-then Boolean logic at its core and looks for mathematical consensus to guide decisions. Mixing these two disciplines can be like expecting oil and water to enjoy each other’s company.
“Relationships and balance sheet still matter, of course,” Greenwich Associates Managing Director Kevin McPartland wrote in a report title that highlights a dichotomy, “Trust and Data Drive Fixed-Income Dealer Growth.”
Trust is a human concept and data is often associated with empirical, mathematically discovered understanding is the point where a melding of the minds might occur. The integration of the two highlights the new path forward. “But the ability to manage both in a more quantitatively driven way can mean the difference between profit growth and a year-on-year decline,” McPartland observed.
Technology is increasingly becoming an important feature on bond trading desks, particularly for sell-side banks. Fully 91% of bulge bracket banks said automating parts of the trading process was one of their “top technology priorities for 2018,” the Greenwich Associates report pointed out. Only 44% of middle market participants, meanwhile, felt the same. This group was having slightly more difficulty “complying with new/changing regulations” and was more focused on improving client management tools and data.
Knowing what technology to integrate so as to deliver a near-term positive return on investment while positioning the firm for a longer-term shift in intelligent automation can prove challenging and, itself, might have variables that don’t easily fit into a math formula.
“Assessing the process of tech integration is always a challenge because no technology is perfect, but it is often much better than human performance,” said Evan Schnidman, CEO and Co-Founder, Prattle, a firm that uses machine learning techniques to analyze human speech patterns to provide a market edge. “Electronic equities trading has proven to be highly reliable, but very rare missteps can have catastrophic cascading effects. With bond trading, and any other technology-driven process in financial services, it is vital to build appropriate circuit breakers and human checkpoints. Simply, technology is a valuable tool, but humans still need to be in the loop.”
When integrating technology into a bond trading workflow it is important to set proper expectations that can be measured on a project plan timeline. Often times expecting the unexpected is a discretionary skill that requires understanding technology’s limits.
“The best practice in tech integration is to remember that technology will not solve all of your problems, but it may be able to streamline the human workflow and optimize decision making,” Schnidman said. “If institutions keep that in mind, they are likely to adopt technology more readily, thereby making each additional innovation less jarring from a tech integration standpoint.”
With technological change coming at an increasingly rapid pace, the imperative is in understand what aspects of human trading can be modeled and what should remain discretionary.
“Instead of fully relying on quantitative approach at the beginning, it is a good starting point to integrate the output of quants and AI models in the human decision making process of position taking and dealing,” said Tsuyoshi Yokokawa, founder and CEO of San Francisco based Alpaca Markets, a firm that in parts helps integrate quantitative trading models into fundamental trading methodologies.
As the last refuge of human-based trading turns algorithmic, the processes and methodologies that drive these projects are likely to determine success or failure to a large degree. In bond trading, the sell side is taking a different approach to that of the middle market.
This article originally appeared on ValueWalk Premium