BCM’s Machine Learning Tactical Model
Last week’s Behind the Markets podcast featured a conversation about tactical asset allocation models using machine learning, with Brendan Ryan, partner and portfolio manager at Beaumont Capital Management (BCM).
BCM believes there are weaknesses in the traditional 60% equity/40% fixed income portfolio approach and believes dynamic and tactical timing strategies can help solve the challenges of today’s markets, like higher equity risk and very low interest rates.
BCM’s separate account strategies started as pure trend-following models that helped navigate the markets through the downturn in 2008. But their models have become more sophisticated, using machine learning algorithms to incorporate more factors in making their tactical decisions.
BCM’s moderate model will typically rotate from a 50/50 stock and bond mix to the high end of 70/30, with 30/70 at the low end, so they are not purely going all in or all out of the market.
The goal of the BCM tactical machine is to find patterns for how investors think and ultimately behave, looking to discover assets that people might ultimately desire in the future.
Loss aversion is one powerful driver of decision-making, and behavioral finance is something BCM incorporates into the machine learning process to figure out what segments of the market to buy.
On the opposite side of the market, BCM’s machine looks for momentum stocks that can run higher for a long time, with technology stocks being the prime example today.
In terms of value, Ryan pointed to the oil sector as today’s prime example, where the flood of supply and lack of demand in the coronavirus pandemic has pushed down energy prices. But these dynamics also should lead to rationing of the oil supply, and the BCM model looks at the energy sector as a potentially good short-term value play, while technology is more of a long-term trend to ride.
For those looking for dynamic and tactical asset allocation models, this was a good primer on how models are becoming more sophisticated and adaptive than the old trend-following models.
You can listen to the full conversation with Brendan Ryan below.