Artificial Intelligence in Equity and Crypto Markets: Progress, Profitability Evidence, and the Limits of Automated Investing
cs.AI · q-fin.PM · q-fin.TR
Quick verdict
Critical review: real AI progress upstream, but no public evidence of persistent net alpha; balanced and technically literate, though non-systematic.
Summary
This critical state-of-the-art review examines public research through 31 August 2026 on AI across listed equities/ETFs, centralized crypto spot, perpetual futures, and on-chain markets, asking whether technical progress translates into sustainable net alpha. The authors organize evidence via an 'alpha-translation chain' (point-in-time information to signal to portfolio to execution to risk-adjusted net returns) and a seven-dimension evidence profile distinguishing capability demos, task metrics, historical out-of-sample tests, costed simulations, prospective portfolios, and audited live capital. Their verdict is deliberately bounded: the record shows real progress in prediction, decision-aware portfolio learning, text extraction, and agentic workflow integration, but evidence thins sharply downstream — look-ahead corrections can erase celebrated alphas, LLM memorization contaminates historical tests, prospective AI portfolios show no significant abnormal returns, and AI-hedge-fund outperformance decays. No general AI architecture is shown to deliver persistent, cross-regime, capacity-aware net alpha. Main limitations: the review is non-systematic with unauditable study selection, its verdict bar is partly circular by its own admission, and pivotal evidence rests on preliminary 2026 working papers.
Review complete
Assessment at a glance
4
risk flags identified
● high 0
● medium 2
● low 2
4 strengths · 5 weaknesses · 5 author questions
AI text signal = Binoculars zero-shot detector score over 20 sampled passages (Fast-DetectGPT co-reported) — absence does not prove human authorship, and a flag does not imply misconduct.
Levels reflect verification complexity — not a quality score. Evolve never rates your paper.
Key claims under review
- Public evidence through 31 August 2026 documents real but mainly upstream AI progress in return prediction, text processing, portfolio design, and workflow integration across equities and crypto.
- Within that public record, no general AI architecture is shown to deliver persistent, cross-regime, capacity-aware net alpha.
- Temporal contamination, repeated selection, survivorship, weak benchmarks, costs, venue mechanics, and capacity routinely break translation from predictive gains to net alpha — e.g., a look-ahead correction can erase a celebrated ML alpha (Zhang et al. 2025).
- Crypto markets require stratum-specific P&L treatment (spot vs perpetual vs on-chain) because funding, liquidation, custody, gas and MEV change the profit equation beyond any equity-style cost scalar.
- Credible future claims require point-in-time data and model checkpoints, decision-aligned objectives, joint portfolio–execution evaluation, prospective protocols, and authority-matched governance.
Strengths
Reusable audit framework, not just commentary
The alpha-translation chain (Eq. 1), seven-dimension profile E=(T,S,P,I,R,X,O) (Eq. 2), and CAP–EXT archetypes (Table 1) operationally separate a good benchmark from deployable net alpha; the CAP/TASK vs NET/PROSP/LIVE distinction directly explains why LLM-agent demos (§5.3) are not profit evidence.
Symmetric treatment of contested evidence
§2.4 and §4.1 hold the replication disagreement open (Hou et al. 2020: most of 452 anomalies fail; Chen & Zimmermann 2022: 319 characteristics largely reproduce; Jensen et al. 2023: international confirmation) and balance cost views (Novy-Marx & Velikov 2016 vs Frazzini et al. 2018) instead of cherry-picking a narrative.
Market-strata P&L literacy
Disciplined secondary reporting
Weaknesses
Verdict partially circular
Load-bearing sources are preliminary
Unauditable study selection
Self-citation inside the evidence table
Thesis never quantified on its own corpus
Full Review Report
$1.99 · one-timeUnlock the complete preliminary analysis Evolve produced for this paper.
Report preview
Scope & Claims
This is a critical state-of-the-art review (explicitly not a systematic review, §2.1) of public research through 31 August 2026 on AI for investing across four strata: listed equities/ETFs, centralized crypto spot, perpetual futures, and on-chain/DEX markets. The central claim (§8.3, §10) is deliberately bounded: the public record shows real but 'upstream' progress — cross-sectional prediction (Gu et al. 2020), decision-aware portfolio learning (Simon et al. 2026; Jensen et al. 2026; Cong et al. 2026), text extraction (Lopez-Lira & Tang 2026; Kelly et al. 2026), and agentic workflow integration (§5.3) — but 'no general AI architecture is shown to deliver persistent, cross-regime, capacity-aware net alpha.' The contribution is the audit framework itself: an alpha-translation chain (Eq. 1) with named failure channels per hop, a seven-dimension evidence profile E = (T,S,P,I,R,X,O) (Eq. 2), and eight evaluation archetypes CAP→EXT (Table 1) that stop 'out of sample' from being conflated with point-in-time validity or with live capital.
The remaining sections cover Methodology Assessment, Statistical Evidence, Results & Reproducibility, Limitations & Risks, and Verdict & Recommendations.
- All 4 strengths & 5 weaknesses, analyzed in detail
- 4 risk flags with severity, section & reviewer questions
- Methods & statistical evidence assessment
- 5 questions a peer reviewer would ask
- Full six-dimension evaluation report & recommendations
One-time purchase · this paper only
One-time purchase, this paper only. AI-generated preliminary analysis — not peer review.
Frequently asked questions
Does this paper prove that AI cannot beat the market?
No. It argues only that the public research record through August 2026 does not demonstrate any general AI architecture delivering persistent, risk-adjusted returns after costs across market regimes. It explicitly leaves open that proprietary systems or strategy-specific approaches may be profitable, stating that public insufficiency is not evidence of private absence.
What is the alpha-translation chain?
The paper's core framework: investment information must survive a sequence of conversions — point-in-time availability, a stable signal, a feasible portfolio, executable orders, and risk-adjusted returns after costs — before it counts as sustainable net alpha. Progress at one stage (e.g., better forecasts) can be destroyed at any later stage by turnover, borrow constraints, spreads, capacity limits, or signal decay.
Is this a systematic review or meta-analysis?
No — the authors explicitly decline those labels in Section 2.1. It is a critical narrative review with iterative search, no registered protocol, no database-complete retrieval, and no pooled effect sizes; they argue quantitative pooling would be misleading because reported Sharpe ratios differ in frequency, annualization, universe, benchmark and cost assumptions.
What does the paper conclude about crypto trading with AI?
Crypto is split into three strata — centralized spot, perpetual futures, and on-chain markets — each with distinct failure modes: wash-traded volume and custody risk, funding and liquidation mechanics, and gas, slippage and MEV respectively. The paper finds structured return regularities exist, but says the AI-specific crypto profitability evidence is thinner than the equity record.
About this review
This is an AI-generated preliminary review by Evolve.Pub, produced from the openly available text of the paper. It is not peer review and does not represent the authors' views. The canonical record of the paper is at the publisher link above.
- Tables and Figure 1 were extracted as flattened run-on text, so exact multi-line table cells (Tables 1, 3, 4, 5) were reconstructed from context; individual cell attributions carry minor misreading risk, though all quoted numbers were verified verbatim in the text.
Go deeper on this paper
Wrote a paper? Stress-test it before reviewers do.
Upload your manuscript for the same structured scrutiny — reasoning, statistics, evidence, figures, and reproducibility.