Over the past decades, asset pricing research has identified an extensive array of stock return predictors and model-based strategies, many reporting substantial improvements in predictive performance. However, predictive accuracy alone does not guarantee investment success. While most studies that examine market frictions focus on trading costs, Álvaro Cartea, Mihai Cucuringu, Qi Jin and Jiexiu Zhu, authors of the November 2025 study “Bottom-up Capacity Constraints and the Limits of Anomaly Profitability,” focused on liquidity constraints — using a stock’s average daily volume (ADV) as a proxy for liquidity to analyze whether strong predictive signals actually translates into profitable trading strategies. Their findings reveal a stark disconnect between academic research on stock return anomalies and their real-world implementability.
What The Researchers Examined
The research team proposed a novel framework to evaluate trading strategies by measuring their implementable value in dollar terms rather than just percentage returns. They applied this approach to 128 well-documented stock return predictors (anomalies) and several machine learning models that combine these predictors. Their analysis included all common stocks publicly traded on the New York Stock Exchange, American Stock Exchange and Nasdaq and spanned the period 1993-2023.
The key innovation was adopting a “bottom-up” approach that estimates tradable dollar volume for each individual stock based on its liquidity constraints, specifically using average daily volume as a proxy for how much capital can realistically be deployed.
The researchers evaluated strategies across two periods:
* In-sample (IS): The original time period used in academic studies that first documented each anomaly.
* Out-of-sample (OOS): The period following initial publication.
Key Findings: A Reality Check For Quantitative Strategies
1. Liquidity Constraints Devastate Profitability
The most striking finding: liquidity constraints alone — before accounting for any trading costs — substantially eliminate profitability for most single-anomaly strategies. Over 70% of these strategies generated less than $1,000 in daily profit during the out-of-sample period. “Even under a highly optimistic scenario (already largely unrealistic) that allows single-stock trades of up to 10% of their ADV and with a $10 million capital cap, the realized returns of only 19 anomalies are statistically significant.”
When liquidity constraints were applied, realized Sharpe ratios declined sharply relative to conventional portfolio benchmarks. In the in-sample period, the average Sharpe ratio fell by 43%, and only about half of anomalies showed statistically significant returns. Out-of-sample, just 20 of 128 predictors remained statistically significant.
2. The Predictability-Profitability Divergence
Perhaps most counterintuitive: many well-known predictors like momentum exhibited strong out-of-sample predictive power yet produced low realized Sharpe ratios and negligible dollar profits when liquidity constraints were considered.
Why? The strongest predictive power was concentrated in illiquid small-cap stocks where tradable volume is severely limited. Meanwhile, most profit originates from large-cap stocks where predictability is weaker and less stable over time.
3. Machine Learning Helps, But Not Enough
The researchers tested various prediction models including penalized linear regression, tree-based methods and ensemble approaches. While these models substantially enhanced predictive performance consistent with prior research, the improvement in actual profitability proved modest.
The gains remained concentrated among small-cap stocks, and improved forecast precision didn’t translate into meaningful increases in realized dollar returns.
4. Trading Costs Are Secondary To Liquidity
When the team incorporated trading costs using standard estimates, profitability declined further. However, liquidity constraints proved to be the primary friction limiting trading capacity. Only 20 anomalies delivered significant out-of-sample returns with liquidity constraints alone; this dropped to 15 with price impact costs and just 11 with bid-ask spread costs.
5. The Fund Size Dilemma
In a case study examining different fund sizes, the researchers observed a troubling trade-off: as investment capital increased, absolute dollar profits rose but the realized Sharpe ratio declined — consistent with well-documented diseconomies of scale in asset management.
Small, nimble funds can target stocks with attractive expected returns, but as funds grow, capital must be diverted into a broader set of less profitable but more scalable securities.
Their findings led the authors to conclude: “The results reveal a clear divergence between predictability and profitability. “They added: “When a strategy targets illiquid stocks, the size of its implementable value is severely limited.” And finally, they noted: “When the fund is small and nimble, it can invest capital into stocks with attractive expected returns. As the fund grows and liquidity constraints start to bind, cash is diverted into a broader set of securities where the additional stocks are less profitable but more scalable.”
Key Investor Takeaways
1. Evaluate strategies by implementable value, not just statistical significance as many anomalies exhibited severe liquidity constraints that limited their economic significance. A high t-statistic or impressive back test Sharpe ratio may be largely unachievable in practice. Always assess:
* How much capital can actually be deployed?
* What are the realistic dollar profits, not just percentage returns?
* Where is the predictive power concentrated — in liquid or illiquid stocks?
2. Small-cap anomalies are statistical mirages for institutional capital. Beware strategies heavily weighted toward micro-caps. If a strategy’s top holdings are concentrated in very small companies, question whether it can scale beyond modest capital amounts. While small-cap stocks may exhibit stronger return predictability, their limited liquidity makes them unsuitable for strategies requiring scale. Focus on improving predictability where capital can actually be deployed — among large-cap stocks.
3. Don’t conflate predictive accuracy with economic value. Machine learning models may boost R-squared and portfolio Sharpe ratios in back tests, but these improvements often don’t translate proportionally to tradable profits. Ask whether enhanced predictions actually increase implementable scale.
4. Consider liquidity constraints before trading costs. Most academic studies focus heavily on transaction costs, but this research shows liquidity constraints impose the binding constraint for many strategies. Evaluate both together rather than in isolation.
5. Small is sometimes better. The research confirms that smaller portfolios can be more nimble, targeting higher-conviction positions that larger funds cannot access efficiently. This is one genuine advantage individual investors maintain over institutions.
The Bottom Line
This research delivers a sobering message: the finance literature’s “factor zoo” of 300+ documented anomalies contains far fewer genuinely profitable strategies than commonly believed. Strong predictive signals don’t automatically translate into investable strategies, especially at institutional scale.
For both research and practice, the framework provides a more realistic assessment of economic value by showing that predictive accuracy alone is insufficient — strategies must also be evaluable by their implementable scale and achievable profits under realistic liquidity conditions.
The gap between what looks good on paper and what works in practice has never been clearer. Before deploying capital based on any anomaly or prediction model, investors should ask not just “Does it predict returns?” but “Can it be traded profitably at the scale needed?”
Larry Swedroe is the author or co-author of 18 books on investing, including his latest Enrich Your Future.
