Understanding the past can sometimes help us better predict what might happen in the future. But in complex adaptive systems “ such as financial markets “ simply understanding the past is usually not good enough.
Regression-based models for predicting stock market behaviour are flawed and no longer deliver the best results.
In the current challenging macroeconomic environment, banks must turn to the latest technologies in order to produce more accurate simulations of future market behaviour.
Why historical data belongs in the past
Ultimately, in the real world, every trade has an effect on the market which changes the behaviour of other traders. Historical data cannot account for this behaviour because it is simply a static recreation of the past.
As a result, it is often unclear if your data has a blind spot in relation to a period of high volatility until it is too late.
Occurrences such as flash crashes and unprecedented market patterns are rare, but that is little consolation if you miss one because your model is unable to predict events that have not happened before, or worse, if your model reacts in a way that contributes to the disorder.
What’s more, historical data is usually expensive to acquire and can be difficult to clean to the point where it is useful.
Pressure to change
When it comes to attracting order flow, sell-side execution desks face increasing competition in an environment in which it is becoming increasingly difficult to differentiate themselves.
The challenge for brokers is navigating a changing regulatory landscape while demonstrating a consistent, reliable performance across a range of market conditions for clients.
Added complexity comes from the regulatory responsibility of preventing algorithms from disrupting the market and having a plan in place to detect and contain those that might.
There is also an element of self-preservation in this in that an unprecedented market event could turn into an existential threat if a firm is reliant on an algorithm that misinterprets it or does not account for it at all.
The role of advanced simulation
In order to offer more competitive trade pricing as well as managing regulatory pressures, banks are turning to advanced simulation via agent-based modelling.
Agent-based modelling is superior to relying on historical data because it enables businesses to simulate the behaviour of individual traders and other agents in the market, thus allowing them to model the systemic effects of specific events or trends.
How does the activity of high-frequency traders affect pricing for a particular group of stocks? What happens if an exchange alters its execution policies? How will a firm’s infrastructure respond to stress scenarios?
Being able to run the same simulation thousands of times, with and without an order, enables banks to compare scenarios and reach an assessment of the likely effect of a trade. This simulation can be run over and over again alongside different execution strategies to determine the one that is most effective.
Bringing the future to life
Previously, banks had to rely on models built from historical data because the computer power needed for more reliable simulation was too great.
But the price of computers has decreased in recent years while the processing power most companies can deploy has increased, making advanced simulation a feasible and reliable reality for many for the first time.
The result is a more accurate estimate on the cost of a trade and the ability to mitigate far better against unintentional herding behaviour, while demonstrating to regulators that you are not contributing to disorderly markets.
Ultimately, understanding the past will only get you so far. The brokers that thrive will be the ones that can bring to the future to life.
