It’s exciting and it’s advancing fast: machine learning in trading

Advanced application of machine learning in trading is part of a trend driving an increasing number of capital markets firms to adopt ML and other AI techniques to build algorithmic trading systems that learn from data without relying on rules-based systems.

In recent years, for example, the number of buy-side clients at American multinational investment bank, JP Morgan, seeking algorithmic execution has soared.

It has expanded from a small number of large hedge fund clients to the more traditional managers, including pension funds, asset managers and insurance companies.

“Five years ago, there were pockets of interest in executing this way, depending on the specific trader or firm’s appetite.

“It’s now become far more mainstream, driven by broader electronification in fixed income markets as well more investment firms adopting more explicit execution benchmarks,” Peter Ward, global head of futures and options electronic execution explained.

The motivation? The period of intense volatility in 2020 due to the global pandemic has played a key role in the increasing buy-side adoption of futures algos as traders became more accustomed to on-screen execution and liquidity.

With the involvement of more data scientists, developments in cloud computing, and access to open source frameworks for training machine learning models, AI is revolutionising the trading desk. Already the largest banks have rolled out self-learning algorithms for equities trading.

Many experts believe that models built with machine learning are faster, more complex, and can synch to extreme events, such as the volatility unleashed by the recent pandemic.

Rather than hard-coding rules in, ML means the system can determine what’s happening in the environment and then inject this new data from the market into its decision calculation.

This ability is being stretched to its limit with research underway by the Oxford-Man Institute of Quantitative Finance.

Home to many of the world’s sharpest mathematicians and data scientists, the institute was founded by the London-based hedge fund, Man AHL, which has $127 billion AUM.

The firm plans to incorporate the research ideas produced at the institute. And now Man Group Plc-backed researchers at the University of Oxford have released news of a machine-learning programme that can project how share prices move.

It is still at the testing stage. But it has, they say, achieved an 80% success rate for the equivalent of about 30 seconds of live trading.

“In the multi-step forecasting, we effectively have a model which is trained to make a forecast at a smaller horizon,” has said Stefan Zohren, an associate professor at the institute who co-authored the research.

“But we can feed this information back into itself and roll forward the prediction to arrive at longer-horizon forecasts.”

So what’s next for machine learning? Experts now have the price volatility of digital assets such as bitcoin in their sights.

Academic studies are underway worldwide using predictors to forecast bitcoin returns such as economic policy uncertainty, equity market volatility index, S&P returns, USD/EURO exchange rates, oil and gold prices, volatilities and returns.

Unlike the stock market, which has almost 100 years of data to study, the cryptocurrency market has only a decade or so of history. So, unsurprisingly, thus far the results have been mixed and inconclusive.

The research is ongoing. The more work put into collecting data and understanding the market, the faster the day will come when there will be predictive tools for cryptocurrency investors.

And that may well mean that ML’s ability to power a model with the predictive power to beat the market may not be so very far away…