Commodities

Executing a Quantitative Investment Strategies (QIS) index based on machine learning

Macquarie and Protean Capital LLP (Protean) execute the first Quantitative Investment Strategies (QIS) index that uses signals based on reinforcement learning for systematic volatility strategies in the market.

Swipe for more
Sector Commodities, Technology
Sub-sector Quantitative Investment Strategies
Location United Kingdom

Opportunity

Protean, a UK fund manager, was looking to add an FX volatility carry strategy to their portfolio.

This strategy sells FX options with delta hedging until expiry to harvest the difference between the implied and realised volatility, i.e. the ‘volatility carry’. In many market conditions, including sideways markets, it can provide an attractive source of return. However, it is not without some downside risk. 

Approach

Macquarie’s Commodities and Global Markets Group leveraged our pioneering technology platform, which incorporates state-of-the-art machine learning models, to develop a reinforcement learning (RL) algorithm that sizes the amount of options to be sold as part of the volatility carry strategy. 

The RL model learns market behaviour through trial and error using feedback from its own trading actions during the model training period. The real skill of applying machine learning in this context lies in the training of the model, by Macquarie, to ensure it takes the most relevant information and successfully encapsulates the current trading and market dynamics. This ensures it is as effective as possible in generating the best trading decisions in live scenarios, where the resulting signals can be easily interpreted by the team.

Outcome

Macquarie was able to further enhance what are already performant strategies through RL techniques by improving the indices’ downside risk profile and risk adjusted returns, as measured by a 0.2 increase in the long-term Sharpe ratio, which is also accompanied by a reduction in other drawdown metrics.

This application of RL in volatility strategies paves the way for its wider use in other systematic investment strategies, as Macquarie looks to build on the existing knowledge base and partner with clients to explore further opportunities. 

Protean have been working with Reinforcement Learning techniques to improve client portfolio returns for a number of years but had found it difficult to find a partner bank to provide concrete implementations. Macquarie were the only bank that provided a skill set and platform which enabled Protean to fully benefit from these techniques.”

Bob Champney
Managing Partner at Protean Capital LLP

Learn more

Related insights

Explore all insights

Case study

Mitigating market risk for Universal-Investment
Woman controlling temperature with a smart thermostat mounted on the wall at modern apartment. Concept of smart technologies at domestic life. Woman wearing green sweater
Woman controlling temperature with a smart thermostat mounted on the wall at modern apartment. Concept of smart technologies at domestic life. Woman wearing green sweater

Perspective

Evolving with precision: A history of innovation in electric meter solutions

Perspective

The divergent futures for metals markets beyond COVID-19