Simply restaging past events is no longer enough when it comes to accurately predicting how markets will evolve, writes Justin Lyon, CEO of Simudyne.

No one really knows how much value is lost because of poor equity trade execution. Current algorithms, relying on heuristics or machine learning, aim to capture and optimise the execution of trades to minimise market risk, impact and price slippage.

Each evolution of an algorithm relies on meaningful data sets for testing before deploying in the real market. The challenge, however, is finding and interacting with these meaningful data sets.

The simulators based on replays of historical data currently in use are outdated and serve only to reconstruct past events, something further complicated by the fact that any given event in the past may never happen again or in the same way. 

Conventional tools are not enough to simulate today’s interconnected market where every move a trader makes sets off a chain reaction too dynamic to predict. Think of interconnected events branching off in an infinite number of directions. The market is a complex ecosystem that adapts to trading behaviours as they occur, as trades are made and as reactions to potential panic or an upside surprise happens in real time.

The solution is to collect anonymised historical data sets and build on these with simulated data based on emerging modelling techniques. Advancements in technology enable us to bring together two things. The first is data from all kinds of sources – the ‘big data’ that surrounds us – which can be fed into a variety of machine-learning algorithms to generate insight. The second is using next-generation emerging modelling techniques – that is, agent-based modelling that uses insights from the natural sciences and advancements in cloud computing – to more accurately capture and simulate what really happens in the capital markets.

Special agents

The Bank of England has issued a number of papers, and chief economist Andy Haldane has commented on the role of agent-based modelling (ABM) in banking supervision.

Furthermore, considering the Markets in Financial Instruments Directive (MiFID II) and the drive for best execution, there is also interest in experimenting with big data, ABM and computational simulation in the capital markets. 

Robert Barnes, CEO of Turquoise, says that the trading venue could use its MiFID public data to seed a lifelike sandbox reflective of a real equity market that can serve as a simulator tool for its members.

Cloud-based computing power and data storage are cost-effective enough for widespread use with artificial intelligence, simulation and machine-learning technology. For example, there are now simulation engines that are very fast, so we can simulate a continuous trading day at high frequency in only a few minutes or seconds.

Agent-based simulation can express how all kinds of market participants interact with one another and how that interaction could cause specific things to happen to them and to the price movements of their equity trades.

A realistic simulation, a distant cousin to the kind popularised in the Matrix series of films, would capture traders’ unique strategies and the price formation process, so we can actually see prices emerge from the interactions of all agents within the simulated market. It is a micro-based, bottoms-up approach to modelling the limit order book.

A tough problem

But capturing the microstructure and the emergence of the price formation process has historically been really hard to do. It is both a very difficult computational challenge and a very difficult modelling challenge.

Institutions that adopt this approach would be able to counterfactually see what would have happened if they traded differently, before deciding on an execution strategy. Traders can visualise market impact by simulating the likely effect of their trades in a dynamic, responsive and high-fidelity simulation of a financial market.

Importantly banks, investors and regulators could understand what would have happened if different strategies were used, and which one would work best in a particular market condition by rejecting the vast majority of the other strategies available.

By taking an ABM approach we can capture the behaviour and unique strategies of primary market participants: noise traders, fundamental traders, momentum traders and market makers. We then bring them together in a simulated market, or sandbox, to understand the best way to execute any equity trade.

By updating their approach to equity trading and algorithmic execution, funds, institutions and banks may be able to gain a significant competitive edge and capture value otherwise left on the table.

In the battle to create more investor value, embracing this scientific approach could yield dividends yet to be discovered.

Justin Lyon is CEO of simulation software company Simudyne.

PLEASE ENTER YOUR DETAILS TO WATCH THIS VIDEO

All fields are mandatory

The Banker is a service from the Financial Times. The Financial Times Ltd takes your privacy seriously.

Choose how you want us to contact you.

Invites and Offers from The Banker

Receive exclusive personalised event invitations, carefully curated offers and promotions from The Banker



For more information about how we use your data, please refer to our privacy and cookie policies.

Terms and conditions

Join our community

The Banker on Twitter