Paper: arXiv 2601.11958

Authors: Zefeng Chen, Darcy Pu

Abstract

Can fully agentic AI nowcast stock returns? We deploy a state-of-the-art Large Language Model to evaluate the attractiveness of each Russell 1000 stock daily, starting from April 2025 when AI web interfaces enabled real-time search. Our data contribution is unique along three dimensions. First, the nowcasting framework is completely out-of-sample and free of look-ahead bias by construction: predictions are collected at the current edge of time, ensuring the AI has no knowledge of future outcomes. Second, this temporal design is irreproducible – once the information environment passes, it can never be recreated. Third, our framework is 100% agentic: we do not feed the model news, disclosures, or curated text; it autonomously searches the web, filters sources, and synthesises information into quantitative predictions. We find that AI possesses genuine stock selection ability, but only for identifying top winners. Longing the 20 highest-ranked stocks generates a daily Fama-French five-factor plus momentum alpha of 18.4 basis points and an annualised Sharpe ratio of 2.43. Critically, these returns derive from an implementable strategy trading highly liquid Russell 1000 constituents, with transaction costs representing less than 10% of gross alpha. However, this predictability is highly concentrated: expanding beyond the top tier rapidly dilutes alpha, and bottom-ranked stocks exhibit returns statistically indistinguishable from the market. We hypothesise that this asymmetry reflects online information structure: genuinely positive news generates coherent signals, while negative news is contaminated by strategic corporate obfuscation and social media noise.

Complexity vs Empirical Score

  • Math Complexity: 3.0/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Street Traders — practical and empirical, lighter on theory

Why this score: The paper presents a highly novel, fully agentic LLM nowcasting framework with strong empirical rigor, including risk-adjusted alphas and transaction cost analyses. Mathematical complexity is relatively low as it relies on standard asset pricing factor regressions rather than theoretical derivations. Reproducibility is limited by design due to the real-time, irreproducible nature of live web search querying, despite the provided repository link.

Research Flowchart

  flowchart TD
    A[Research Goal: Test Agentic AI's Ability to Nowcast Stock Returns] --> B[Input: Daily Russell 1000 Universe]
    B --> C[Agentic AI Process: Autonomous Web Search, Filtering & Information Synthesis]
    C --> D[Out-of-Sample Predictions: Daily Rank & Attractiveness Scores]
    D --> E[Trading Strategy: Long Top-Ranked Stocks vs. Short/Hold Others]
    E --> F[Key Outcome: Top 20 Stocks Yield 18.4 bps Daily Alpha & 2.43 Sharpe Ratio]
    E --> G[Key Outcome: Predictability Asymmetry - Broad Portfolios Dilute Alpha, Short Side Ineffective]

Cited References (77)

  1. Efficient capital markets: A review of theory and empirical work (1970)
  2. The efficient market hypothesis and its critics (2003)
  3. Some anomalous evidence regarding market efficiency (1978)
  4. The limits of arbitrage (1997)
  5. Limits of arbitrage (2010)
  6. Presidential address: Discount rates (2011)
  7. Common risk factors in the returns on stocks and bonds (1993)
  8. Taming the factor zoo: A test of new factors (2020)
and 69 more
  1. Dissecting characteristics nonparametrically (2020)
  2. Arbitrage portfolios (2021)
  3. Predicting returns with text data (2019)
  4. ChatGPT and corporate policies — → 2409.17933
  5. Sentiment during recessions (2013)
  6. Wisdom of crowds: The value of stock opinions transmitted through social media (2014)
  7. Is all that talk just noise? The information content of internet stock message boards (2004)
  8. Textual analysis in accounting and finance: A survey (2016)
  9. The rise and potential of large language model based agents: A survey (2023) — link
  10. Can ChatGPT beat the stock market? Testing limits of LLMs on trading (2023)
  11. AutoGen: Enabling next-gen LLM applications via multi-agent conversation (2023) — link
  12. ExpeL: LLM agents are experiential learners (2024)
  13. A five-factor asset pricing model (2015)
  14. On persistence in mutual fund performance (1997)
  15. Returns to buying winners and selling losers: Implications for stock market efficiency (1993)
  16. Transaction costs of factor-investing strategies (2021)
  17. Open source cross-sectional asset pricing (2021)
  18. Trading costs (2018)
  19. Investor sentiment and the cross-section of stock returns (2006)
  20. The sum of all FEARS investor sentiment and asset prices (2015)
  21. International stock return predictability: What is the role of the United States? (2013)
  22. Stock return predictability and variance risk premia: Statistical inference and international evidence (2014)
  23. Machine learning panel data regressions with heavy-tailed dependent data: Theory and application (2022)
  24. Nowcasting: The real-time informational content of macroeconomic data (2008)
  25. Characteristics are covariances: A unified model of risk and return (2019) — link
  26. Factors that fit the time series and cross-section of stock returns (2020)
  27. Empirical asset pricing via machine learning (2020)
  28. Deep learning in asset pricing (2024) — link
  29. Forest through the trees: Building cross-sections of stock returns (2025)
  30. (Re-)Imag(in)ing price trends (2023) — link
  31. Machine learning from a ``universe’’ of signals: The role of feature engineering (2025)
  32. Machine learning vs. economic restrictions: Evidence from stock return predictability (2023)
  33. Artificial intelligence asset pricing models (2025)
  34. Giving content to investor sentiment: The role of media in the stock market (2007)
  35. When is a liability not a liability? Textual analysis, dictionaries, and 10-Ks (2011)
  36. News implied volatility and disaster concerns (2017)
  37. The structure of economic news (2023)
  38. Business news and business cycles (2024)
  39. Can ChatGPT forecast stock price movements? Return predictability and large language models — → 2304.07619
  40. Expected returns and large language models (2022)
  41. Bloated disclosures: Can ChatGPT help investors process financial information? (2024)
  42. Financial statement analysis with large language models — → 2407.17866
  43. Discretionary disclosure (1983)
  44. Annual report readability, current earnings, and earnings persistence (2008)
  45. Have capital market anomalies attenuated in the recent era of high liquidity and trading activity? (2014)
  46. Does academic research destroy stock return predictability? (2016)
  47. Algorithmic trading and market quality: International evidence (2021)
  48. … and the cross-section of expected returns (2016)
  49. Does algorithmic trading improve liquidity? (2011)
  50. NLP evaluation in trouble: On the need to measure LLM data contamination for each benchmark (2023) — link
  51. Time travel in LLMs: Tracing data contamination in large language models (2023) — link
  52. Measuring massive multitask language understanding (2021) — link
  53. TradingAgents: Multi-agents LLM financial trading framework — → 2412.20138
  54. Empirical asset pricing with large language model agents (2024) — link
  55. FinGPT: Open-source financial large language models — → 2306.06031
  56. BloombergGPT: A large language model for finance (2023) — link
  57. InstructFinGPT: Financial sentiment analysis by instruction tuning of general-purpose large language models — → 2306.12659
  58. TradingGPT: Multi-agent system with layered memory and distinct characters for enhanced financial trading performance — → 2309.03736
  59. FinAgent: A multimodal foundation agent for financial trading — → 2402.18485
  60. Learning to generate explainable stock predictions using self-reflective large language models — → 2402.03659
  61. FinMem: A performance-enhanced LLM trading agent with layered memory and character design — → 2311.13743
  62. LLM-Factor: Extracting profitable factors for stock prediction from news using large language models (2023) — link
  63. Emergent abilities of large language models (2022)
  64. Sparks of artificial general intelligence: Early experiments with GPT-4 (2023) — link
  65. A theory of financial media (2022)
  66. Do managers withhold bad news? (2009)
  67. Anomalies and financial distress (2016)
  68. Efficient capital markets: II (1991)
  69. High frequency market microstructure (2015)

10 of 77 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.