Fixed Income & FX Leader Summit APAC 2019

17 - 19 September, 2019

The Westin, Singapore

65 6722 9455

Embracing Machine Learning And AI
In Fixed Income Trading

Technologies like AI (artificial intelligence), machine learning (ML), and process automation have the potential to revolutionize fixed income trading. Significant disruption has already occurred, with markets in the more developed economies having had the benefit of using such cutting edge technologies to increase liquidity, transparency, and the ease and speed of trade over the past few years.

As fixed income traders contemplate the possibility of trusting vital trading decisions to a machine, the time is right to consider the implications of widespread deployment of AI and its associated technologies in the global marketplace.

Fixed Income Trading At A Crossroads

At the Fixed Income Leaders Summit at Amsterdam in 2018, buy-side speakers remarked upon the increasing emergence of automated processes in fixed income trading, and the resulting need for buy-side desks to adapt in order to keep pace with these developments. New services and new ways of finding liquidity based on software and digital processes are changing the workflows of human traders, who must either embrace these challenges or seek employment elsewhere.

AI, Intelligence, And The Question Of Trust

The complex algorithms and multi-market handling capabilities of high-frequency trading platforms are no longer at the cutting edge of fintech. HFT is giving way to AI, and platforms powered by even more complex mathematics and the potential for autonomy. AI systems with deep learning capability are able to detect and reason out best outcomes, based on the amount and type of data they absorb - at least, to a certain extent. Over time, and with augmented data sets, their ability to make context-based decisions can improve to levels that will make them viable consultative "partners" to fixed income traders.

But the question arises: How can investors trust such a system?

img

Organizations like Google and IBM have made strides in developing some AI techniques based on deep learning that can accumulate the current active price of a bond from various sources. But it's early days, and these are relatively immature technologies with little background to show that they have been performing consistently, or to prove that the new strategies employed in their development have been working.

And, given that AI has yet to gain universal acceptance, there's been little coordination among banks and brokers over the standardization of the new technology, and the conditions of its use for fixed income trading.

Machine Learning, Or Humans Guiding?

Much of the debate centers on the degree of automation and autonomy that trading systems based on AI and machine learning should have. In order to embrace deep learning and AI for enhanced alpha generation and improved workflow efficiencies, it's necessary to identify which aspects of the trader workflow best lend themselves to automation - and where you need to retain the human touch.

For fixed income trading, this balancing act between automation across the desk and human intervention also requires making decisions about high touch and low touch operations. With advances in technology and higher buy-side expectations fueling a shift towards low-touch trading, organizations must develop a best-fit execution policy for automation, defining a set of criteria for low touch trading that still allows traders to focus on the high touch business.

img

Challenges And Opportunities

Artificial intelligence derives from the power of the algorithms and automated responses underlying it - and the reliability of the data to which it has access. The "garbage in, garbage out" principle carries even greater weight in a system whose capacity to learn depends on the quality of information that it absorbs. The configuration of AI systems must therefore incorporate tools or processes to clean and clarify data inputs, and establish criteria for their degree of relevance. The fixed income trading platform developed by Overbond gives an illustration of the level of design that's required.

img

(Image source: Overbond)

Methods available for computers to explain the logic behind their decisions to humans are still in their infancy. As yet, even the best software is unable to dictate an investor’s preference for risk, or intuitions about specific opportunities. Machine learning models are also limited in their ability to predict the activities of governments and central banks that affect the markets, especially during exceptional circumstances. So the AI - human intelligence balance must still be factored into the mix of any deployment.

Traders looking to adopt machine learning or AI platforms must recruit, train, and retain the appropriate human talent to handle transition and deployment of the new technology. For example, while high-touch specialist traders may have a lot of experience in their field, they may have difficulty getting involved with the new technology that’s on offer to them. Change management and corporate culture shifts are also required at a wider scale, to instill the correct behavioral adjustments in a buy side organization.

Issues Of Adoption

Challenges aside, adoption of AI and machine learning for fixed income trading has nonetheless been proceeding at a healthy pace. In its inaugural 2019 Artificial Intelligence / Machine Learning Global Survey, Refinitiv found that financial institutions have gone beyond experimenting with and testing machine learning, to actively applying machine learning models. Refinitiv expects AI to be the single greatest enabler of competitive advantage in the financial sector.

Global adoption figures for machine learning put the Asia Pacific region at the heart of this trend, with 76% of those institutions polled considering ML as a core component of their business strategy.

img

(Image source: Fintech Singapore)

This enthusiasm is understandable. AI, process automation, and machine learning offer significant potential in providing advanced analytics such as price discovery, liquidity risk management, intelligence gathering, pre-trade and post-trade analytics for fixed income trading and other financial activities.

Download the agenda today for more information and insights.

img