How we seek to enhance our alpha model

At MDT Advisers, we are systematic, quantitative investors; underlying all of our work is our alpha model. What underlies the alpha model itself? Intuition and research.

Our model didn’t arrive fully formed like Athena, of course, nor will it remain fixed. Instead, we use a rigorous research process to review the model’s output. The goal is to deliver sustained investment outcomes by improving the individual alpha factors, risk controls, portfolio construction and even the modeling process itself.

Some of those updates have been notable, such as when we enhanced our decision tree framework to a “forest” of decision trees. Some have been below the surface, but all have been in service of that goal.

A look under the hood

What’s the process like? We have a six-month research cycle. It’s a disciplined process, but it starts in an organic way. We begin with a research brainstorming meeting, incorporating ideas from academic journals along with internal analysis and observations made by the investment team. Additionally, we bring in observations gleaned from our quarterly performance attribution meetings and from our daily trade reviews, where we examine trades and the news flow to see if something is happening that our model could better capture. Is there, we ask, a characteristic of companies that we want to study that the current model doesn’t account for?

Taken together, this all gets combined into a brainstorming document where team members pitch potential ideas to research. From this document, we develop an agenda where the most promising ideas, with the clearest means of investigating them, are studied.

Not just any idea makes it onto the agenda. An idea needs to be specific, plausible, and actionable. It must identify an area of the investment process that is worth improving and have a logical plan for how to address it. In other words, we have to have some hope of getting there from here. If there are too many steps along the way, or if the way is too hazy, then the idea may be too big or too vague. In that case, we would work to refine the idea first, perhaps breaking it down into smaller, more feasible steps.

The six-month process allows innovation to be regularly implemented without constant upheaval. When I started here as a research analyst, the research cycles lasted two years. The conclusion of a cycle could prompt portfolio turnover in excess of 10%. The shorter cycles we now use are designed to be more streamlined and less disruptive.

Along the way, we try to develop an account of what types of improvements we think might be achievable. It’s helpful to have a narrative of areas of potential improvement to work from since the possible set of topics to investigate is otherwise too vast.

As for the research itself, we’ll take data and test hypotheses against the data, using backtest simulations that we strive to make as accurate as possible. Does the idea demonstrate consistent improvement through entire market cycles, in different economic environments, and across different cap ranges? Does the idea interact with other parts of the process in ways that align with our expectations? As for that data, we like it like sushi: as raw as possible. We like to say that “we don’t buy signals from vendors.”

Aesthetics and investing

When we find an idea that works, we then try to simplify it to its essence. It may be surprising to learn that quantitative investing requires a significant amount of aesthetic sense. We try to make each factor as simple and elegant as possible. Often, the sign of a good idea is that there is a spot where it seems to fit perfectly, and it seems obvious in retrospect.

My team is forever on a six-month treadmill. Is that a dreary feeling? Hardly. Each time we find something new. Like Sisyphus, in Albert Camus’ retelling, “The struggle itself toward the heights is enough to fill a man’s heart. One must imagine Sisyphus happy.”