Quants like boasting that sophisticated models power their trading strategies, not humans. But the models, it turns out, can be as varied as humanity.

Take value, a pillar of factor investing that can be distilled as “buy low, sell high.” Brainchild of Nobel laureate Eugene Fama and Ken French, it’s a simple enough concept. The trouble comes when you need to decide what’s low and high.

There are no fewer than 3,168 ways to do it, according to a tally by a team of quant researchers led by Morgan Stanley’s Stephan Kessler for a new study published in the Journal of Portfolio Management.

“There is no consensus around what value actually is,” wrote the authors, who also include Bankhaus Lampe KG’s Bernd Scherer and University of Wuppertal’s Jan Philipp Harries. “Taking Fama and French’s work as a point of departure, researchers face many (active) decisions when constructing a value portfolio.”

The paper is more than mere academic wonkery. Its findings strike at the heart of the debate over quantitative investing, which promises scientific rigor but whose reliance on human decision-making can spur sharp distinctions within a single style. It’s especially relevant for value at a time when the factor’s decade-long underperformance has led some to claim it’s structurally broken.

“These results underline further the challenges in defining value in a narrow sense and ascribing certain characteristics to the style,” according to the authors.

The researchers found that most metrics deliver stronger gains than the classic but controversial definition used in the Fama-French model, which looks at book-value-to-price. Focusing on cash flow and earnings delivered the best risk-adjusted returns, while using dividend yields delivered the worst.

The gap between the best- and worst-performing value strategies over nearly three decades was as much as 463 percentage points, according to the study.

The quants found that value’s returns are mostly driven by long rather than short bets, echoing recent research by Robeco. Kessler and company also reveal that when it comes to shorting, it’s better to bet against indexes than individual names. Removing sector biases—which happen when certain industries like banking become cheap on the whole—can help reduce drawdowns, they write.

The pitch for factor investing is that academically-sourced drivers of returns—such as quality, value and momentum—can be systematically harvested without much human skill. But as the paper suggests, human decision-making, with all its biases and blindspots, is everywhere. And they can profoundly influence returns.

Apart from picking a signal, humans decide how to weight each stock. They have to choose whether to short an index or individual names, how to treat sector biases and when to re-balance the portfolio, according to the paper, which looked at shares in the S&P 500 Index.

“Strategy labels such as value, momentum, or quality may be misleading classifications because they suggest a larger commonality among strategies,” according to the authors.

This article was provided by Bloomberg News.