This paper looks at about 200 published “anomaly” trading strategies. An anomaly here means a rule that sorts stocks into winners and losers and then buys the winners and shorts the losers. Measured across all stocks and using the historical periods reported in the original papers (through December 2005 and earlier), the median long-short return was about 48 basis points per month (0.48%). But when the same set of strategies is evaluated only after 2005, the median falls to 19 basis points per month. If the authors also restrict attention to more investable stocks—the roughly 3,000 largest stocks that together make up about 90% of market value—the post‑2005 median falls to about 7 basis points per month. Even small trading costs or allowances for chance would eliminate returns of that size.

To reach these numbers the authors use an open, standardized replication dataset from Chen and Zimmermann (2022) that reproduces nearly all published cross‑sectional return predictors. They split the sample into two eras—through 2005 and post‑2005—to capture changes in market structure with the rise of modern trading technology. They also impose simple “investability” filters (for example, keeping the top 3,000 stocks or the top 90% of market capitalization) to approximate what a non‑micro‑cap portfolio manager could actually trade. Sparse or tiny portfolios are treated conservatively so the post‑2005 results reflect strategies that could be implemented in real time.

The paper also looks at statistical significance and the role of luck. Each strategy’s t‑statistic is the usual measure of whether its average return is unlikely to be due to sampling noise. Across the roughly 200 strategies, the cross‑sectional variance of t‑statistics after 2005 is near what pure luck would produce. Using a shrinkage adjustment that separates signal from noise, the authors find that most of the apparent returns collapse toward zero. For example, the single raw top performer in their post‑2005, investable universe had a raw average of 66 basis points per month but a luck‑adjusted return of only about 6 basis points.