Artificial Intelligence in Equity and Crypto Markets: Progress, Profitability Evidence, and the Limits of Automated Investing

Battle Count: 50

This paper is directly and comprehensively relevant to quantitative trading. It provides a critical evaluation framework (alpha-translation chain) for assessing whether AI/ML methods actually produce sustainable net alpha in equity and crypto markets. It covers the full investment workflow: data ingestion, prediction, portfolio construction, execution, and risk management. It addresses market microstructure, transaction costs, capacity constraints, and implementation issues that are central to quantitative trading. The paper evaluates specific trading strategies, agent systems, RL policies, and portfolio optimization methods. It provides a structured evidence profile and evaluation-design archetypes directly applicable to quantitative strategy development and validation. The crypto-specific analysis (perpetual futures, on-chain markets, MEV, AMMs) is highly relevant to crypto quantitative trading. The paper's central finding—that no general AI architecture is shown to deliver persistent, cross-regime, capacity-aware net alpha—is a critical caution for quantitative traders.

This is a review/synthesis paper, not an implementation paper. It does not propose a specific model or system to implement. The implementation complexity score reflects that the paper itself requires no implementation. However, the frameworks it discusses (alpha-translation chain, 7-dimension evidence profile, evaluation-design archetypes, market-specific P&L equations, governance architecture) are conceptually complex and would require significant engineering effort to operationalize in a trading system. The paper's value is in providing evaluation criteria and research directions rather than a deployable solution.

This is a critical review paper, not an empirical study with a single reproducible pipeline. It synthesizes evidence from hundreds of cited studies. The review itself provides a structured framework (alpha-translation chain, 7-dimension evidence profile, evaluation-design archetypes) that is conceptually reproducible. However, the paper explicitly states it is NOT a systematic review: it does not claim database-complete retrieval, duplicate independent screening, a registered protocol, or formal risk-of-bias scores. Two ancillary audit aids (claim-evidence map and representative study-level evidence profile) are included with the source package but remain selective aids. The paper emphasizes that many cited studies themselves lack reproducibility due to undisclosed data, proprietary models, and incomplete reporting.