Daniele Grassi, CEO of Axyon.AI
From the shifting economy to the changing political landscape – everything has an impact on the market. Yet, even with all the advancements in econometrics and analysis over the last few decades, the sector has still been unable to fully anticipate the true effect these changes will have.
The answer lies in the past and understanding how to harness historical incidents to inform future events. Those firms that can unlock the data hidden in these trends will be able to lead the market, see increased profits and avoid the negative impact of market shifts.
The challenge of looking back
There is an innate challenge in using past events to anticipate future change, however. Markets face extremely complex dynamics where thousands upon thousands of actors interact in non-linear ways. As a result, it becomes extremely complicated to identify the underlying trends and processes that will influence future change. However, through understanding these intricate relationships, firms can begin to anticipate regime shifts and even black swan events.
It’s here that the traditional quantitative analysis, supported by human supervision, is not enough. There is simply too much data and too many variables to consider. Even if analysts do make predictions, these are mainly based on surface-level trends. However, there are many more patterns and dynamics that traditional quantitative analysis and the human mind are unable to perceive, such as small shifts in entirely different industries that can create huge waves for the entire market.
A storm in a clear sky
The benefits of anticipating market shifts go beyond remaining competitive or increasing profit; these insights can also offer protection against entirely unexpected developments. While experts may be able explain in hindsight how a black swan event occurred, this understanding is not always enough to prepare or offset the negative impact of future events.
Every time a black swan occurs, the circumstances causing it appear to be different. Nevertheless, the result always ends up disrupting and damaging many investors and institutions.
Part of the reason for this lack of preparation is due to how firms currently adapt to market change. Traditional stress tests are not perfect, as they tend to replay a limited set of scenarios in order to predict how future market changes could affect a portfolio or the risk balance of an institution in general.
Even when these historical scenarios have more complex variations included, they often use a limited range of variables that do not account for deeper market shifts. This is where scenario analysis needs a more effective process.
With the proper tools, human investigations can be enhanced to provide more accurate predictions of the market. However, technology is often still a stumbling block for many financial institutions, due to a fundamental lack of understanding and whether it will deliver practical benefits for the business.
The reality is far more positive, as AI solutions can actually help workers to be more efficient in their role. As a result, financial professionals should welcome the use of technology to help predict future trends and market issues.
New machine learning techniques, such as Generative Adversarial Networks (GANs), can support this goal by modelling the market with far greater accuracy. The data used in the formation of these scenarios can be more detailed and broader in scope, which means that the technology can look at data points coming from a range of actors that influence the market – such as economic data, fundamentals, sentiment and news. GANs can then use all of this nuanced information generate scenarios that take into account the inner workings of the market and not just surface-level events.
In practice, this is achieved by having two AIs working against one another. One AI produces fake scenarios while the other decides whether that data is real or false. As the AI learns to spot the false data, its counterpart improves its practices to make the next set of data, or market scenario, even more realistic. By using this kind of synthetic data, potentially in the form of thousands and thousands of years of realistic market scenarios, the planning and preparation for market changes can be constantly developed and not limited to static events that have taken place in the past.
While currently existing at a research-level, this technology is likely to reach mainstream adoption in the next few years. If firms are prepared to make this transition, they will be able to gain a far better understanding of how the market will shift, not only by drawing directly from historical events, but also by understanding how variations on these events can shape the market as well.