Stock market anomalies, which by their nature perform contrary to the notion of efficient markets, are appealing to investors.

They’re also risky because they are mercurial—rising, falling and vaporizing, seemingly on a whim.

New research by Nejat Seyhun of the University of Michigan Ross School of Business and others decodes some of the myths and mysteries of anomalies from an investment perspective.

Seyhun and colleagues looked at the role of the value factor, which is measured by the market-to-book ratio of a given stock. The book value is the value of a company as stated in its balance sheet and the market value is derived from the value of its stock.

So, if a stock has a high market-to-book ratio, it is considered expensive, and a low ratio is considered cheap.

The researchers investigated the interaction between 13 well-known anomalies and the value factor, and found the factors are most predictive when the stocks that they contain are also cheap. When expensive, the factors don’t seem to predict the future stock returns very well.

Specifically, Seyhun says they found anomalies that exhibit a value (cheap) orientation outperform anomalies that exhibit a growth (expensive) orientation by about 30 basis points per month. He adds that cheap anomalies plus favorable recent stock price movements outperform expensive and unfavorable anomalies by about 90 basis points per month (or 11% a year).

Alternatively, more than 96% of the dollar return for the anomalies disappears when they have negative-momentum and expensive orientations.

Among the more well-known anomalies is the “January effect.” The month long has been known for having higher-than-average stock-market returns, but does it? Turns out the anomaly was strong in the 1970s and ’80s but has mostly declined—and even disappeared for some years—since the ’90s.

The study notes that growing numbers of investors exploiting the same anomaly can lead to an anomaly strategy becoming more expensive in terms of valuation—leading to a deterioration in future returns.

The finance industry has been highly productive in discovering hundreds of so-called anomalies, says Seyhun, professor of finance and business administration. While that’s exciting, he cautions that “so many predictive factors make a mockery of the concept of market efficiency, which states that stock prices reflect all publicly available information.”

“Our research suggests that it’s OK to ignore these anomalies and focus on what academic research suggests works over the long-term as starters: market exposure, book-to-market ratio and firm size,” he said. “Momentum also appears to be useful as a timing factor.”

The study is published in the Journal of Empirical Finance. Co-authors include Deniz Anginer of Simon Fraser University, Sugata Ray of the University of Alabama and Luqi Xu of the College of Charleston.