This Tuesday’s roundup highlights the most useful and actionable investment insights I came across over the past week, drawing from academic papers, industry research, blogs, and thoughtful discussions on social media. Links to all original sources are included throughout.
Commodities
Commodity Futures and the Limits to Arbitrage (Cooper, Day, Lewis, and Molyboga)
Commodity index investing appears to have changed the economics of futures curves. Since financialization took off after 2003, index commodities developed significant term premiums, while premiums in non-index commodities are not statistically significant. The premiums also increase with maturity and passive index participation. Key takeaway: Passive investment flows can systematically influence expected returns across the futures curve.
Equities
Retail Trading and the Varying Predictive Power of Aggregate Short Interest (Froymovich, Gong, Wasley, and Xiao)
One of the strongest historical predictors of market returns has become much weaker. Aggregate short interest’s ability to predict future S&P 500 returns declined substantially from 2012 to 2023, alongside a sharp rise in retail trading. The authors link this decline to more retail short selling and greater short-squeeze risk. Key takeaway: The predictive value of market signals can change as the investors behind them change.
Binary Bias and Stock Returns (Fattinger, Hanspal, Koval, and Steshkova)
Stocks with a high frequency of down days over the previous month subsequently outperform those dominated by up days. A long-short strategy earned 12.3% annually from 1963–2023, and the effect is not explained by conventional short-term reversal. Key takeaway: The frequency of up vs. down days predicts returns beyond the magnitude of recent price moves.
Does Weinstein Stage Analysis Beat a Moving Average? (Roskill)
Stan Weinstein’s four-stage framework is widely used in technical analysis, but its edge may come almost entirely from simple trend following. Testing Weinstein’s method on survivorship-free S&P 500 data since 1992, the author finds that its useful information comes from whether the 30-week moving average is rising, not from identifying bases and tops. Key takeaway: The trend signal works; the additional stage classification in Weinstein’s framework adds little forecasting value.
Quantifying the Contributions of Clustering to Statistical Arbitrage (Zhu, He, and Cucuringu)
How you group stocks may matter more than the algorithm used to group them. Across 27 years of U.S. equities, peer-based statistical arbitrage produces higher alpha and substantially lower portfolio volatility than no clustering or arbitrary groups. Yet 12 very different clustering methods lead to surprisingly similar portfolios. Key takeaway: Much of the value comes from identifying informative peers, not fine-tuning the clustering method.
Intermediate-Term Reversal (Huang, Vincent, and Yeh)
Ranking U.S. stocks by their largest peak-to-trough decline between 12 and 24 months before portfolio formation produces a strong reversal effect: The high-minus-low portfolio earns 1.16% per month, with 0.87% eight-factor alpha. The effect survives among large caps and after controlling for momentum and other characteristics. Key takeaway: The path of past prices contains predictive information that cumulative returns can miss.
ETFs
In Defense of Leveraged and Inverse Funds (Carpenter, Lu, and Whitelaw)
This paper argues volatility decay in leveraged ETFs is widely misunderstood. Daily rebalancing does not inherently destroy value; it creates a convex payoff: Leveraged and inverse ETFs underperform a linear leverage benchmark when index moves are modest, but outperform when moves are sufficiently large. Key takeaway: Leveraged ETF returns are nonlinear by design, not systematically eroded by daily rebalancing.
FX
Taking Carry Trading Seriously (Breedon and Vitale)
The standard “high-minus-low” carry strategy does not capture carry as well as commonly assumed. Using FX data from 1983–2025, the paper constructs an optimized carry portfolio accounting for risk and trading frictions. Its net Sharpe is 0.71 vs. 0.32 for conventional high-minus-low carry, and its returns show little exposure to the risk factors typically associated with carry. Key takeaway: How you construct a carry portfolio fundamentally changes both its performance and sources of risk.
Machine Learning & Large Language Models
Return-Optimal Regime Labels and Progressive Risk Overlays: A Hierarchical Framework for Industry Allocation (Li and Mulvey)
Machine learning can be especially valuable for dynamic risk allocation, not just return prediction. This paper trains an XGBoost model across 46 industry pairs, then combines within-equity rotation with defensive overlays and volatility scaling. Out-of- sample from 1995–2025, Sharpe rises from 0.56 to 1.01 while max drawdown falls from 57% to 22%. Key takeaway: Portfolio construction can be as important as prediction: The largest incremental Sharpe improvement comes from the defensive overlay.
The Return Impact of Form 10-K Releases: Evidence from a Large Language Model (Frey, Qiu, and Zhao)
Markets do not appear to fully digest 10-K filings when they arrive. A fine-tuned LLM predicts returns over the subsequent 20 trading days from the filing text alone. Stocks in the highest predicted-return quintile outperform the lowest by about 5 percentage points, value-weighted. Key takeaway: Long corporate disclosures contain return-relevant information that is not immediately reflected in prices.
Blogs
What are U.S. Treasury markets really telling us? Part II. (Hanno Lustig)
Improving Fixed-Weight Portfolios with Dynamic Rules (Quantseeker)
Can AI Agents Beat the Random Walk? Not So Fast (CFA Institute)
Podcasts
Lucas Schuermann - Swapping Out Perpetual Futures (Flirting with Models)
Hedge Fund Manager Alix Pasquet: How Small Funds Can Outsmart Multi-Managers (Odds on Open)
Why Governments Are Afraid of the Bond Market (Robin Wigglesworth Explains) (Meb Faber)
Zigs, Zags, and Finding Micro Cap Winners with Ian Cassel (RCM Alternatives)
Social Media & Industry Research
Inflation Redux? (AQR)
IPOs of doom (Owen Lamont, Acadian Asset Management)
The Quant Renaissance Part III: Who's Ready for the Next Winter? (Man Group)
Last Week’s Most Popular Links
A Sharpe of 2.1 From Nothing: The Second Number Your Agent Doesn’t Log (Jonathan Kinlay)
Rethinking Predictability: Distributional Forecasts of the Market Risk Premium (Sinha)
Correlation and Commodity Risk Premia (Fan and Zhang)
Disclaimer: This newsletter is for informational and educational purposes only and should not be construed as investment advice. The author does not endorse or recommend any specific securities or investments. While information is gathered from sources believed to be reliable, there is no guarantee of its accuracy, completeness, or correctness.
This content does not constitute personalized financial, legal, or investment advice and may not be suitable for your individual circumstances. Investing carries risks, and past performance does not guarantee future results. The author and affiliates may hold positions in securities discussed, and these holdings may change at any time without prior notification.
The author is not affiliated with, sponsored by, or endorsed by any of the companies, organizations, or entities mentioned in this newsletter. Any references to specific companies or entities are for informational purposes only.
The brief summaries and descriptions of research papers and articles provided in this newsletter should not be considered definitive or comprehensive representations of the original works. Readers are encouraged to refer to the original sources for complete and authoritative information.
This newsletter may contain links to external websites and resources. The inclusion of these links does not imply endorsement of the content, products, services, or views expressed on these third-party sites. The author is not responsible for the accuracy, legality, or content of external sites or for that of any subsequent links. Users access these links at their own risk.
The author assumes no liability for losses or damages arising from the use of this content. By accessing, reading, or using this newsletter, you acknowledge and agree to the terms outlined in this disclaimer.
Paid subscriptions may not be available in all jurisdictions and may change without notice.
On the clustering paper: the finding that twelve fairly different methods converge on similar portfolios matches what I see running a thematic cluster map in the equity book I run. The algorithm is close to interchangeable. What is not interchangeable is the assumption that a peer group keeps behaving like one. The interesting failures are rarely misgrouped names, they are clusters that stop trading as a block, where leadership rotates inside the theme and the peer relationship quietly stops holding. A study measuring the stability of peer sets through time would be more useful to me than another comparison of clustering algorithms.