Weekly Research Recap

Latest research on investing and trading

With 2026 now underway, this week’s recap curates the most actionable investing insights from the past seven days across academic research, industry reports, blogs, and social media, complete with direct links.

---

Commodities

Gold and Silver Shine During the Day (Kelly)

Using hourly sky-cover data from four New York City weather stations, the paper shows that weather-driven mood impacts intraday returns on precious metals. Clear versus cloudy days add 8–9 bps for GLD and 16–18 bps for SLV, with no effect on overnight or daily returns. Key takeaway: Weather-induced mood creates behavioral return patterns in precious metals that appear only during active trading hours.

---

Crypto & Digital Assets

Tokenized Gold (Harvey, Rabetti, Lin, and Zhang)

Using 2020–2025 blockchain transactions and crypto-exchange prices, tokenized gold (PAXG, XAUt) mirrors the economic behavior of physical gold. Prices stay closely linked to spot and futures, with acceptable liquidity and spreads. During periods of macroeconomic and market turmoil, tokenized gold attracts flight-to-safety demand. Key takeaway: Tokenized gold enables continuous, 24/7 trading without altering its risk–return profile.

---

Currencies

Global currency volatility and depreciation factors (Londono and Yamani)

Using FX option data, the paper identifies two priced factors, U.S. dollar volatility (ATM implied vol) and tail risk (10-delta risk reversals). High-minus-low portfolios earn 0.33 to 0.34% per month (Sharpe ratios of 0.5 to 0.6) and raise cross-sectional R² for carry and momentum from 19% to 33–44%. Takeaway: FX options embed global risk premia and partly explain FX carry and momentum while providing standalone return predictability.

---

Equities

Macroeconomic Shocks and Cross-sectional Stock Returns (Fieldhouse, Kim, and Mahajan)

Macroeconomic shocks like monetary, credit, oil, and fiscal shocks causally predict cross-sectional stock returns. Using 1968–2024 data, these shocks explain a sizable share of the variation in factor and anomaly returns. E.g., monetary and credit tightening hits mispricing factors hardest, while oil shocks mainly affect energy industries through cash-flow channels. Key takeaway: Factor returns are highly macro-dependent, suggesting scope for factor timing based on macro conditions.

---

ESG

Revisiting ESG Performance: Do High Scores Translate to Higher Returns? A Risk-Adjusted Analysis of S&P 500 Portfolios (Carvalho, Falcao, Pinheiro, and Carrao)

Using 2005–2024 data, low-ESG portfolios have vastly outperformed high-ESG portfolios in raw returns (923% vs. 89%). Risk-adjusted performance is also higher for low-ESG firms (modified Sharpe about 1.1 vs. 0.5), though the difference is only marginally statistically significant (p-value about 0.09). Key takeaway: So-called “brown” firms have massively outperformed “green” firms in raw returns, although the difference in risk-adjusted performance is marginal in a statistical sense.

---

Machine Learning & AI

When Markowitz Meets Machine: Optimization of Large Portfolios with High-Dimensional Stock Characteristics (You and Zhang)

This paper reframes mean–variance optimization as an end-to-end ML problem that maps firm characteristics directly into portfolio weights. For China A-shares (2017–2022), it achieves an information ratio of 2.40 vs. 1.56 for traditional MVO after costs (CSI 300), and 3.81 vs. 3.09 for CSI 500, helped by lower turnover. Key takeaway: Integrating forecasting and portfolio construction beats forecast-then-optimize.

Generative AI-enhanced Sector-based Investment Portfolio Construction (Voronina, Romanko, Cao, Kwon, and Mendoza-Arriaga)

The paper asks 12 large language models to select and weight 20 stocks per S&P 500 sector and tests the portfolios out of sample. LLM portfolios often outperform in stable markets but underperform in volatile periods; results improve when paired with mean–variance optimization. Key takeaway: LLMs help in stock selection, but are most effective when combined with traditional portfolio optimization frameworks.

Uncertainty-Adjusted Sorting for Asset Pricing with Machine Learning (Liu, Luo, Wang, and Zhang)

Uncertainty-adjusted ML ranking outperforms point-prediction sorting by accounting for stock-specific forecast reliability. Bounds are constructed from quantiles of historical out-of-sample prediction errors around each forecast. In U.S. equities, Sharpe ratios increase significantly; for example, neural-network portfolios improve from 1.48 to 1.86. Key takeaway: Explicitly incorporating prediction reliability when sorting assets improves portfolio performance.

A Test of Lookahead Bias in LLM Forecasts (Gao, Jiang, and Yan)

Much of the apparent forecasting power of large language models may stem from lookahead bias rather than genuine inference. The authors introduce “Lookahead Propensity (LAP)”, a diagnostic that measures how familiar a prompt is to the model. Higher LAP signals a greater risk of bias. Key takeaway: Before trusting LLM-based backtests, ensure the reported “alpha” reflects real-time reasoning, not recalled future information.

Enhancing FX Portfolio Allocations through Reinforcement Learning (Luisi, Wang, Tobek, and Bouloutas)

A reinforcement-learning agent that directly sets FX carry portfolio weights outperforms mean–variance optimization in long-run out-of-sample tests (2005–2024). The RL strategy achieves positive returns and Sharpe ratios, while MVO portfolios are negative, with the strongest relative gains during volatility spikes and regime shifts. Key takeaway: Adaptive, reward-driven allocation can outperform static optimization in FX carry portfolios.

---

Portfolio Construction

Optimal Sharpe Ratio Portfolios and the Role of Perfect Foresight into Returns, Volatilities, and Correlation (Roy, Motson, and Thomas)

Proper modeling of correlations, not return forecasting, is the main driver of Sharpe-optimal performance. Even perfect identification of next year’s top returners yields lower Sharpe ratios because winners tend to move together, hurting diversification. Key takeaway: Superior risk-adjusted returns mainly come from correctly modeling correlations among assets and preserving diversification, not from stock-picking accuracy.

---

Blogs

The Volatility You Can’t See (Concretum Group)

Top 10 Blogs of 2025: Insights on Market Cycles and Financial History (CFA Institute)

Cross-Asset Price-Based Regimes for Gold (Quantpedia)

---

Podcasts

The Paradox of Skill: Why AI Makes Active Investing Harder, Not Easier, with Larry Swedroe (ETFatlas)

Stories sell: managed futures, replication and agency problems, with Andrew Beer (Capital Horizons)

Club Conversation with Howard Marks, Oaktree’s Quiet Legend (MicroCapClub)

---

Social Media & Industry Research

Introduction to Forecast Signals (SystematicLongShort)

The Hidden Risks of Leveraged Single-Stock ETFs (Alpha Architect)

The Woes of A (Portfolio) Manager (SystematicLongShort)

2025 (Ray Dalio)

---

Last Week’s Most Popular Links

On the Anatomy of Trend (Kjaer)

What Successful Investors Read: Book Recommendations from Professionals (CFA Institute)

Daily + 15:45 OHLCV: A Database for Reliable Backtesting (Concretum Group)

---

---

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.