The Gutenberg printing press, invented in the mid-15th century, revolutionised the dissemination of information. It democratised knowledge, reduced the cost of book production, and accelerated the spread of ideas, leading to significant cultural, scientific, and economic transformations. Similarly, AI is poised to revolutionise various sectors by automating complex tasks, enhancing decision-making, and enabling new forms of creativity and innovation.

By Francis Akpata

Since the 2008 global financial crisis, a persistent question has haunted the hedge fund industry: do event-driven managers generate true alpha that justifies their fees? When the S&P 500 returned 32.6% in 2013 while the HFRI Fund Weighted Composite Index delivered just 9.13%, pension funds from CalPERS to state retirement systems announced they would reduce or terminate absolute return allocations, citing complexity, fees, and underwhelming performance. More than a decade later, with comprehensive performance data from 2014-2024 now available, this debate has evolved but remains unresolved. The question is no longer simply whether event-driven funds can beat the market, but whether they serve a more nuanced role as portfolio diversifiers in an era of unprecedented market concentration and regulatory intervention.

This article examines how event-driven hedge fund strategies have evolved from the post-2008 scepticism through today's transformed landscape, integrating historical context with contemporary realities to assess whether these funds still deliver meaningful value to institutional investors.

The Post-Crisis Foundation: 2008-2013

The 2008 financial crisis fundamentally altered the event-driven landscape. These funds—which historically profited from corporate events like mergers, distressed restructurings, and activist campaigns—found themselves operating in a radically different environment. The information edge that once distinguished elite managers began eroding rapidly. Where relationships with investment bankers, lawyers, and brokers had previously provided asymmetric information advantages, technological advancement democratized access to data. Real-time news feeds, algorithmic processing, and computerised trading systems meant that competitors quickly arbitraged away opportunities identified by one fund.

The classic event-driven playbook—taking long positions in acquisition targets and shorting acquirers, exploiting index rebalancing flows, or capitalising on holding company discounts—became increasingly crowded. Market neutrality, the holy grail of absolute returns, proved elusive as correlations across asset classes increased during periods of stress. Managers who had promised 'beta-free' returns discovered that their supposedly uncorrelated strategies moved in lockstep during volatility spikes.

Yet the crisis also created opportunities. Credit markets froze, leaving profitable companies unable to access capital despite strong fundamentals. This capital drought was particularly acute in emerging markets and stigmatised sectors like real estate, where sovereign ratings and negative sentiment obscured genuine value. Event-driven managers who could provide flexible structured finance—mezzanine loans, bridge facilities, collateralised loan obligations—found themselves filling a void that traditional banks could no longer address post-Dodd-Frank.

The period also witnessed a strategic pivot toward activism. Rather than passively betting on announced deals, successful funds began participating actively in corporate governance, pushing underperforming companies to unlock value through cost reduction, capital allocation discipline, or strategic alternatives. This wasn't merely opportunistic—it represented a philosophical shift from information arbitrage to value creation through engagement.

The Modern Landscape: 2014-2024 Performance Reality

The decade following 2013 delivered a verdict on event-driven strategies that is both humbling and nuanced. The S&P 500's historic bull run—driven overwhelmingly by large-cap technology—produced annualised returns of approximately 12-13%. The HFRI Event-Driven (Total) Index, by contrast, delivered 5-7% annually. On the face of it, this appears to validate the critics: why pay hedge fund fees for performance that lags passive index funds by so much?

Yet this surface-level comparison obscures crucial context. During the 2022 bear market, when the S&P 500 collapsed by 18%, the HFRI Event-Driven index declined by just 4%, validating the original promise of these strategies: downside protection and portfolio diversification rather than pure return maximisation. For institutional investors managing multi-decade liabilities, this asymmetric return profile—lower upside capture but significantly reduced drawdowns—justifies continued allocation despite absolute underperformance during bull markets.

The 'easy money' era of 2014-2021, characterised by near-zero interest rates and quantitative easing, proved particularly challenging. When all boats rise with the liquidity tide, the value of sophisticated event analysis diminishes. It was only when conditions normalised in 2022 that event-driven strategies demonstrated their intended purpose—not as growth engines, but as ballast during turbulence.

The Fee Structure Evolution

The traditional '2 and 20' model has fractured into divergent paths. Multi-manager platforms like Citadel and Millennium now employ 'pass-through' fee structures, in which investors pay all operating expenses and bonuses, often totalling 3-5% or more. Paradoxically, institutional investors accept these elevated costs in exchange for consistent, low-volatility returns and professional risk management infrastructure that single-manager funds cannot replicate.

Meanwhile, single-manager event-driven funds face brutal fee compression. The average management fee has declined from 2% to approximately 1.4%, while performance fees have dropped from 20% to 16.5%. This squeeze forces a strategic choice: accumulate assets under management to maintain revenue, or remain boutique and performance-focused. Those prioritising asset gathering inevitably dilute their opportunity set; those maintaining discipline may struggle to sustain operations.

Simultaneously, capital that historically flowed to distressed-debt hedge funds has migrated toward private credit and private equity. These alternatives offer the illiquidity premiums that pension funds need to match long-dated liabilities, plus simplified governance compared to hedge fund boards. As predicted in 2013, many pension funds did indeed reduce hedge fund allocations—not necessarily because the strategies failed, but because alternative vehicles better matched their operational and liability structures.

Regulatory Headwinds and Market Structure Changes

The regulatory environment has fundamentally reshaped event-driven tactics. SEC Rule 13D modernisation in 2023, effective 2024, represents the single most significant change for activist investors since the original rule's adoption. The disclosure window for 5% stakes was shortened from 10 days to 5 business days, eliminating much of the 'ambush' advantage where activists could quietly accumulate 9% or larger positions before the market knew. The initial stock pop upon disclosure—historically a primary profit driver—now materialises over a compressed timeframe, reducing expected returns on activist campaigns.

The SEC's adoption of Rule 13f-2 in 2023, requiring more frequent and granular short position reporting, introduced new risks for merger arbitrage and distressed strategies. Funds betting on deal breaks or bankruptcy scenarios now face increased 'short squeeze' vulnerability, as their positions become more visible to retail and institutional counterparties seeking to engineer pain trades.

The rise of aggressive FTC enforcement under Lina Khan's leadership transformed merger arbitrage from a relatively predictable spread-capture strategy into a high-stakes bet on antitrust litigation outcomes. Deals like Microsoft/Activision and Kroger/Albertsons faced unprecedented regulatory scrutiny, widening spreads and increasing volatility. For managers who could correctly forecast legal outcomes, this created opportunity; for others, it introduced deal-break risk at levels not seen since 2008.

The Meme Stock Factor and COVID Disruption

The 2021 GameStop saga introduced a risk factor that did not exist in 2013: coordinated retail liquidity swarms targeting institutional short positions. The collapse of Melvin Capital—though primarily a long/short equity fund—demonstrated that 'valid, thesis-driven' short positions could be overwhelmed by sentiment-driven buying, regardless of fundamental analysis. Event-driven funds using shorts to hedge merger arbitrage or distressed exposures found themselves vulnerable to this new dynamic, in which social media coordination could force the liquidation of otherwise sound positions.

The COVID-19 pandemic created both a crisis and an opportunity. March 2020 saw merger arbitrage spreads blow out to levels unseen since 2008 as deals were delayed or renegotiated, testing managers' conviction and risk management. Simultaneously, the pandemic generated a short but intense distressed cycle in travel, hospitality, and energy sectors where funds like Apollo and Oaktree deployed capital at attractive entry points. The swift recovery, aided by unprecedented fiscal and monetary stimulus, compressed this distressed opportunity window far more rapidly than historical cycles.

Strategy Evolution: Adapting to the New Reality

From Information Edge to Analytical Superiority

Where the 2013 analysis identified eroding information advantages, 2024 reveals a complete transformation. Alternative data has become table stakes—credit card transaction data, geolocation analysis of retail parking lots, web-scraping for pricing changes—all are commoditised inputs. The edge no longer comes from possessing this data, but from deploying artificial intelligence and machine learning models to interpret it faster and more accurately than competitors.

Natural language processing now parses thousands of pages of bankruptcy filings or merger agreements instantaneously, identifying termination clauses, material adverse change provisions, and regulatory hurdles that previously required armies of junior analysts. This technological leverage enables funds to analyse more opportunities with greater depth, but also means that mispricings are corrected more rapidly. The half-life of any identified edge has collapsed from weeks to hours or even minutes.

Portfolio Construction and Liquidity Management

The evolution in portfolio construction reflects lessons learned from multiple crisis episodes. The 2013 observation that diversification asymmetrically protects—with high downside correlation but low upside correlation—has been validated repeatedly. Successful managers now construct portfolios with explicit scenario analysis across deal outcomes, regulatory decisions, and market regime changes, rather than relying on historical volatility measures or value-at-risk models.

The tension between liquidity and performance remains central. Funds that offer quarterly or even monthly redemptions struggle to hold positions through the natural maturation of event-driven investments. The most successful practitioners have embraced the 2013 recommendation to control fund liquidity, implementing longer lock-ups or side pockets for illiquid positions. This aligns manager and investor interests, reduces leverage requirements, and enables concentration in high-conviction opportunities without the distraction of forced selling driven by redemptions.

Geographic Reorientation: From China to Japan

The 2013 thesis identified Eastern Europe, Asia, and African frontier markets as fertile ground for event-driven strategies, where less efficient markets and looser regulations created exploitable buyer-seller imbalances. The subsequent decade partially confirmed this thesis, but with a dramatic geographic reallocation.

China—once viewed as a promising venue for activism—has mainly become uninvestable for event-driven strategies due to geopolitical tensions, regulatory unpredictability, and governance concerns. Funds have systematically reduced their exposure to China and Hong Kong, abandoning what appeared in 2013 to be a generational opportunity.

Conversely, Japan has emerged as the premier event-driven market globally from 2019 to 2024. Tokyo Stock Exchange reforms demanding improved capital efficiency, combined with corporate governance changes and cultural shifts toward shareholder value, created ideal conditions for activism. Elite funds, including Elliott Management and ValueAct, launched high-profile campaigns at Toshiba and Seven & i Holdings, achieving outcomes that would have been unthinkable in the previous generation. South Korea's parallel 'Value Up' initiatives suggest similar opportunities may be emerging across developed Asia, representing a fundamental reorientation from the emerging market focus of 2013.

The Alpha Equation Updated: Redefining Success

Assessing whether event-driven funds justify their fees today requires abandoning simplistic performance comparisons against the S&P 500. The appropriate framework recognises that these strategies serve multiple portfolio functions:

• Asymmetric downside protection: The 2022 performance differential (HFRI ED -4% vs S&P -18%) demonstrates that during market dislocations, event-driven strategies fulfil their diversification mandate, even if they underperform during bull markets.

• Liquidity provision: Event-driven managers continue to reduce corporate cost of capital by providing sophisticated structured financing when traditional sources withdraw, particularly valuable during credit contractions.

• Corporate governance enhancement: Activist campaigns, despite regulatory headwinds, still force management accountability and capital discipline at underperforming companies, generating positive externalities beyond fund returns.

• Volatility-adjusted alpha: While absolute returns trail equities, risk-adjusted performance often compares favorably when accounting for lower volatility and reduced drawdowns.

The most evident failure has been in managing expectations. Event-driven funds never were, and never will be, pure alpha generators that beat equity indices across all market regimes. Their value proposition is lower-volatility absolute returns with equity-like performance during stress. Measured against this standard, many funds have succeeded, even as they disappointed investors, benchmarking them against the S&P 500's extraordinary decade.

The fee question remains contentious. Single-manager funds, compressing management fees to 1.4% and performance fees to 16.5%, better align with documented value creation. Multi-manager platforms charging 3-5% total must demonstrate consistent risk-adjusted outperformance to justify their premium. The migration to private credit structures suggests that for specific strategies—particularly distressed and structured credit—alternative vehicles may offer superior economics for both managers and investors.

Future Trajectory: The Path Forward

Several forces will shape event-driven strategies over the coming decade. Continued technological advancement—particularly in AI-driven analysis and alternative data—will further compress the window for exploiting mispricings, favouring firms with superior computational infrastructure. This creates natural advantages for multi-manager platforms and well-capitalised single-manager funds, while putting pressure on smaller operations.

Regulatory evolution remains a wildcard. While current SEC rules have curtailed some activist tactics, political cycles could reverse these headwinds. More fundamentally, increased antitrust enforcement—whether it persists or moderates—will continue to influence merger arbitrage risk premiums and opportunity sets.

The geographic reorientation toward Japan and, potentially, Korea suggests that developed-market activism, rather than emerging-market opportunism, may dominate the next phase. This reflects a maturation of the strategy—less frontier exploring, more systematic exploitation of structural inefficiencies in sophisticated markets where regulatory frameworks support shareholder engagement.

Most critically, the bifurcation in strategy between liquid and illiquid vehicles will accelerate. Traditional hedge fund structures offering quarterly liquidity increasingly struggle to compete with private credit funds, special-situations vehicles, and closed-end structures that can hold positions through complete event cycles without redemption pressures. The 2013 insight that liquidity management determines performance has proven prescient—funds that controlled redemptions outperformed those that prioritised asset gathering.

Conclusion: Finding Alpha in the Evolution

The evolution from 2013's post-crisis scepticism to 2024's transformed landscape reveals that event-driven hedge fund strategies remain viable, but with fundamentally altered value propositions. The information advantages that once defined success have been replaced by technological and analytical superiority. Pure market-neutral absolute returns have proven elusive, but asymmetric downside protection delivers meaningful portfolio value. Fee structures have bifurcated into premium-priced multi-manager platforms and compressed-fee single-manager operations, each serving distinct investor needs.

The 'alpha in an event' exists not as a permanent edge, but as a temporary dislocation that sophisticated managers can identify and exploit before markets correct. Success requires continuous adaptation—embracing technology, managing liquidity deliberately, maintaining focus rather than diversifying excessively, and accepting that outperformance during market stress may matter more than capturing the whole upside during bull runs.

For institutional investors, the lesson is nuanced: event-driven strategies justify allocation not as equity replacements, but as diversifiers that reduce portfolio volatility and provide downside protection during dislocations. Those expecting these funds to beat the S&P 500 annually will remain disappointed. Those recognising their role as ballast during storms—accepting lower absolute returns in exchange for reduced drawdowns—may find them indispensable components of sophisticated portfolios.

The fundamental question posed in 2013: Do hedge funds generate alpha?—persists, but the answer has evolved. They develop a specific type of alpha: risk-adjusted, downside-protected returns that emerge most clearly during volatility rather than persistent market-beating performance. Whether this justifies fees depends entirely on an investor's objectives, time horizon, and capacity to withstand lower absolute returns in exchange for smoother journeys. The alpha is still there—but finding it requires understanding what one is genuinely looking for.

Francis Akpata

December 2025

The Evolving Relationship Between Human Judgment and Algorithmic Systems in Financial Markets

Francis Akpata  |  Managing Director, Plymouth Capital

Executive Summary

Financial markets stand at a critical juncture. The traditional dominance of human judgment is increasingly challenged by algorithmic systems powered by artificial intelligence and machine learning. This analysis examines that evolving relationship—exploring how techniques from gradient boosting to deep reinforcement learning have transformed market dynamics, while considering where human expertise remains irreplaceable. The conclusion is not one of replacement but of integration: the future belongs to hybrid models that combine human intuition with computational power, governed by robust regulatory and risk management frameworks.

1. The Shifting Landscape of Financial Decision-Making

At every year-end, the investment community undertakes its reckoning: who made the best bets, who identified the mispriced stock, who had the counterintuitive insight that the consensus had missed? For much of modern financial history, that question was answered by human beings—analysts, portfolio managers, and traders — drawing on experience, judgment, and the accumulated pattern recognition of long careers.

That is changing. Sophisticated algorithmic systems now carry out many tasks once reserved for skilled traders, and they do so at speeds and scales that human cognition cannot match. Recent research confirms the acceleration: AI has materially improved predictive accuracy and trading efficiency, reducing human error, latency, and transaction costs (Srivastava et al., 2024). The integration of AI-driven trading has fundamentally altered market dynamics, with broad implications for liquidity, price discovery, and systemic risk (Seth, n.d.).

The central question is no longer whether algorithms can outperform humans in defined tasks—they demonstrably can. The question is whether human judgment retains an indispensable value that machines cannot replicate, and how the two can work together most effectively.

2. The Case for Algorithmic Trading: Speed, Scale, and Objectivity

The most consequential advantage algorithms hold over human traders is the elimination of cognitive and emotional bias. When making investment decisions, humans are subject to the weight of experience, which distorts judgment. An analyst who suffered heavy losses in emerging markets, or who was invested in technology equities just before the dotcom crash, carries those memories into every subsequent assessment. The damage is not merely psychological; it is epistemological. Confirmation bias—the tendency to seek evidence that supports a pre-existing view—is among the most documented and destructive forces in investment decision-making.

Algorithmic systems are immune to these distortions. They track complicated market patterns, analyse real-time data from news and social media, and execute decisions without fatigue, depression, or emotional attachment. AI-enhanced algorithms consistently outperform traditional models in trade decision-making, price prediction, risk assessment, and portfolio management (Srivastava et al., 2024). They operate continuously, monitoring global markets across multiple time zones, identifying arbitrage opportunities, and adjusting positions in milliseconds—capabilities that provide substantial competitive advantages in modern markets (Addy et al., 2024).

3. Machine Learning and the New Frontier

Modern trading algorithms have evolved far beyond simple rule-based programmes. Today's most sophisticated platforms employ advanced machine learning techniques that learn from data, adapt to changing conditions, and discover patterns invisible to human analysts.

Gradient boosting methods, particularly XGBoost, have become foundational in quantitative finance. These ensemble approaches significantly outperform traditional statistical models in stock price prediction and pattern recognition (Leng, 2024). Deep learning architectures—Long Short-Term Memory networks, Convolutional Neural Networks, and Transformer models—have proven effective at capturing temporal dependencies and complex market dynamics. When combined with alternative data such as news sentiment, these models achieve performance that exceeds market benchmarks (Lan et al., 2025).

Perhaps the most significant recent advance is reinforcement learning (RL), which enables trading agents to learn optimal strategies through interaction with market environments. Multi-agent deep RL frameworks have achieved remarkable results, enabling agents to navigate complex market dynamics and adapt to changing conditions (Shavandi et al., 2022). Advanced RL approaches incorporating meta-learning allow trading systems to adapt rapidly to new market regimes by leveraging prior experience (Wang et al., 2025)—a meaningful step beyond the static rule-based systems of an earlier era.

Most recently, large language models (LLMs) have brought natural language understanding to financial markets. BloombergGPT—trained on Bloomberg's extensive proprietary financial datasets—and open-source models like FinGPT enable automated processing of earnings calls, regulatory filings, and financial news at scale (Liu et al., 2023). LLM-driven trading agents such as Trading-R1 combine language reasoning with reinforcement learning, articulating decision-making in natural language and thereby addressing longstanding concerns about algorithmic opacity (Xiao et al., 2025).

4. The Rise of Quantitative Hedge Funds

These technological foundations have produced a new breed of investment firm. Renaissance Technologies—founded by mathematician James Simons—remains the paradigmatic example. Its Medallion Fund has generated average annual returns exceeding 35% after fees over three decades, a performance unmatched in the industry, achieved through sophisticated statistical models, machine learning algorithms, and vast computational resources deployed to identify fleeting market inefficiencies (Fan, n.d.).

Citadel Securities handles approximately 40% of all US retail equity volume; Two Sigma employs hundreds of PhDs in mathematics, computer science, and physics; AQR systematically exploits documented market anomalies—value, momentum, carry, and defensive factors—across global markets. These firms represent the vanguard of a broader shift in which quantitative methods are applied not merely to high-frequency execution but to longer-horizon investment strategies grounded in economic theory.

The competitive advantage of these funds increasingly depends on alternative data: satellite imagery tracking retail traffic, credit card transaction flows, web-scraped pricing data, geolocation signals, and social media sentiment. Machine learning algorithms processing these datasets significantly enhance predictive accuracy (Srivastava et al., 2024). Yet the proliferation of alternative data has intensified competition. As more firms access comparable datasets and employ similar techniques, the predictive power of any single source tends to decay—a phenomenon whose implications we consider in Section 7.

5. The Human Edge: Intuition, Judgment, and Qualitative Analysis

Against this formidable case for algorithmic superiority, the argument for human judgment rests on a set of capabilities that remain genuinely difficult to mechanise.

Algorithms, like economic models, are constructed by human beings with particular perspectives and prescribed objectives. They are powerful within the parameters of their design but brittle at the margins. Human beings are more flexible: they can consciously or subconsciously factor in emotions, irrational sentiments, and the complex texture of human motivation before reaching a decision. A human research analyst spots changeable factors—shifts in tone, ambiguities of intent, the quality of silence in a management call—that computerised systems require time to learn, if they learn them at all.

This qualitative capacity matters most in three domains. First, the assessment of management: human investors can evaluate the competence and integrity of leadership teams through direct engagement—meetings, facility visits, and conversations with customers and competitors—in ways that no natural language processor can fully replicate. Second, the assessment of unprecedented events: when an algorithmic model has been built on certain premises, human beings must intervene when unplanned events occur without historical precedent. The attacks of 11 September 2001, a sovereign debt crisis, or a global pandemic do not feature in training data. Research on interpretable machine learning confirms that human judgment remains essential for interpreting results and adjusting strategy when market regimes shift in structurally novel ways (Zhang, 2025). Third, the understanding of human irrationality itself: experienced investors possess an intuitive theory of mind—an understanding of how fear, greed, overconfidence, and herding manifest in prices—rooted in shared psychology and social experience that machines can model statistically but cannot inhabit.

Recent research exploring whether LLMs can overcome behavioural biases suggests that AI systems may help investors avoid cognitive traps (Liu et al., 2024), but the question of whether machines can fully replicate human insight into market psychology remains open. The most effective approach may be one that combines algorithmic objectivity with human understanding of behavioural dynamics.

6. Flash Crashes and the Systemic Risks of Automation

The increasing dominance of algorithmic trading has introduced new and distinctive forms of market instability. On 6 May 2010, US equity markets experienced a sudden, catastrophic collapse: within minutes, the Dow Jones Industrial Average plunged nearly 1,000 points—approximately 9% of its value—before recovering most of those losses in the same session. Nearly $1 trillion in market capitalisation evaporated temporarily.

Investigation revealed that a large sell order executed algorithmically interacted with high-frequency trading systems, amplifying movements and creating a liquidity vacuum. As prices fell, algorithmic systems began withdrawing or selling aggressively, triggering a self-reinforcing downward spiral (Nahar et al., 2024). The event was not isolated: flash crashes have recurred in US Treasury markets, currency pairs, and individual equities with increasing frequency. High-frequency financial market simulations using agent-based modelling illuminate how these scenarios emerge from the interaction of multiple algorithmic strategies—revealing market fragmentation and heightened volatility as inherent features, not aberrations, of algorithm-dominated markets (2022).

The bigger systemic risk is one of correlated strategies. When multiple market participants employ similar algorithmic approaches based on comparable data and machine learning techniques, hidden correlations accumulate. Crowded trades create systemic vulnerabilities: simultaneous unwinding of correlated positions can trigger cascading failures. The competitive dynamics of quantitative finance incentivise the adoption of successful strategies, accelerating convergence and increasing fragility. Research on AI-driven risk control mechanisms highlights the need for monitoring systems that detect dangerous correlations before systemic crises develop (Vance et al., n.d.).

The opacity of machine learning systems compounds these challenges. Deep neural networks and ensemble methods may achieve superior performance, but their decision-making processes are often opaque even to their designers. This complicates risk management, regulatory oversight, and the explanation of decisions to clients and counterparties—creating a fundamental tension between model complexity and the interpretability that sound governance requires (Zhang, 2025).

7. Alpha Decay and the Innovation Imperative

Alpha decay refers to the erosion of a trading strategy's excess returns as more market participants discover and exploit the same inefficiency. Research confirms that the half-life of trading signals has shortened dramatically as markets have become more competitive: strategies that once generated alpha for years may now lose effectiveness within months (Tang et al., 2025). The arms race in alternative data acquisition reflects this dynamic—datasets that confer significant edge when first discovered become commoditised within a year or two as competitors gain access.

This creates a relentless innovation imperative. Firms must continuously develop new strategies, explore novel data sources, and adopt emerging technologies to maintain competitive advantage. The AlphaAgent framework—which employs LLM-driven alpha mining with regularised exploration specifically designed to counteract decay (Tang et al., 2025)—exemplifies the frontier of this race. Yet as AI-driven strategy discovery becomes widespread, there is a real possibility that alpha decay will itself accelerate: the arms race consuming the advantage it seeks to preserve.

The limiting factor is ultimately human: the intellectual capital of the mathematicians, physicists, and machine learning researchers who develop novel algorithms and extract signals from complex data. Talent competition has driven compensation to extraordinary levels. The concentration of this capital at a small number of leading firms explains their persistent performance advantages and constitutes the most durable moat in an industry otherwise characterised by rapid imitation.

8. Hybrid Human-AI Investment Processes

The most promising path forward lies not in choosing between human and algorithmic decision-making but in combining their complementary strengths. Successful hybrid systems typically employ a tiered architecture: algorithms handle routine analysis and execution while human managers focus on higher-level strategy, qualitative assessment, and intervention during anomalous conditions.

In practice, this means algorithms screen thousands of securities, identify potential opportunities, and execute trades, while portfolio managers make final decisions on position sizing, risk limits, and strategic allocation. Research on multi-agent LLM frameworks demonstrates that collaborative systems combining multiple AI agents with human oversight achieve superior outcomes (Xiao et al., 2024). LLM-driven systems with interpretable reasoning exemplify the augmented intelligence paradigm: providing human investors with AI-generated insights and recommendations while maintaining human decision-making authority (Xiao et al., 2025).

Human intervention remains structurally important even in automated markets. The Brady Commission's introduction of circuit breakers—mechanisms that halt trading when indices fall below defined thresholds—represents an early institutional acknowledgement that computerised systems require human guardrails. As market stress increases or unusual patterns emerge, human involvement must intensify. The design of effective adaptive oversight frameworks, capable of scaling human engagement to market conditions in real time, is among the most pressing practical challenges facing the industry.

9. Regulatory Oversight and Governance

The transformation of financial markets by algorithmic trading has prompted regulatory responses aimed at maintaining market integrity and mitigating systemic risk. Frameworks such as MiFID II in the European Union and consolidated audit trails in the United States require firms to register algorithmic systems, maintain detailed trading records, implement risk controls, and provide regulators with access to algorithms for inspection (Nahar et al., 2024). MiFID II imposes specific organisational requirements on algorithmic traders, including pre-deployment testing, real-time monitoring for disorderly trading, and robust business continuity arrangements.

Yet traditional regulatory approaches face genuine challenges when applied to AI-driven systems. The opacity of machine learning models makes it difficult for regulators to assess compliance or systemic risk. The adaptive nature of reinforcement learning systems means that algorithms may evolve in ways not anticipated by their developers. Regulators are increasingly leveraging AI themselves to improve market surveillance—deploying monitoring systems capable of detecting anomalous trading patterns and potential manipulation (Vance et al., n.d.). Even so, the rapid evolution of AI techniques and the proprietary character of trading algorithms make comprehensive systemic risk assessment extremely difficult.

The potential for RL trading agents to engage in implicit collusion or manipulation without explicit programming raises novel challenges: when agents discover that certain behaviours increase profits, they may converge on strategies that, while not intentionally manipulative, produce manipulative effects (Jafree et al., 2025). The global nature of algorithmic trading compounds these difficulties, as regulatory arbitrage enables firms to locate operations in jurisdictions with lighter regulatory oversight. Effective oversight ultimately requires international coordination and principles-based regulation focused on outcomes—market integrity, investor protection, systemic stability—rather than prescriptive technical requirements that may quickly become obsolete.

10. Conclusion: Toward a Balanced Future

Applying a quantitative, computerised approach to investing produces a measurably clearer understanding of the investment process—eliminating cognitive, emotional, and confirmation bias, and enabling more precise outcomes. This is a genuine and substantial advance. Yet it does not replace fundamental analysis, economic assessment, common sense, or, most importantly, a human being's intuitive understanding of other human beings in the market.

The evolving relationship between human judgment and algorithmic systems is not a story of displacement but of transformation and integration. Algorithms have demonstrated clear superiority in processing vast datasets, identifying complex patterns, executing with precision, and maintaining the objectivity that emotional biases deny human traders. Large language models represent the latest frontier—bringing natural language understanding and reasoning to financial analysis in ways that will further alter the landscape.

Yet the limitations of algorithmic systems remain significant and, in some respects, structural. Flash crashes demonstrate that systems optimised for normal conditions behave unpredictably under stress. Model risk and systemic risk grow as markets become more dependent on AI decision-making. Alpha decay ensures that competitive advantages prove fleeting. And human judgment retains enduring value precisely where algorithms are weakest: navigating unprecedented events, assessing qualitative factors, understanding the psychology of other market participants, and recognising when fundamental assumptions have changed.

The future of financial markets will not be decided by the triumph of algorithms over human intelligence, or vice versa, but by our collective ability to design systems that combine the best of both—the speed, scale, and objectivity of machines with the intuition, judgment, and adaptability of the human mind. In that balanced future, the question ceases to be 'man versus algorithm' and becomes, instead, how man and algorithm can work together to create markets that are more efficient, more stable, and more fair.

References

Addy, W. A., Mensah, I. K., & Ofori, E. K. (2024). Algorithmic trading and AI: A review of strategies and market impact. World Journal of Advanced Engineering Technology and Sciences, 11(1), 054.

Dung, N., Chen, Y., & Wang, L. (2025). SAFE machine learning in quantitative trading. https://doi.org/10.2139/ssrn.5015984

Fan, L. (n.d.). The rise of AI quantitative investment funds.

Jafree, M., Chen, X., & Liu, Y. (2025). When AI trading agents compete: Adverse selection of meta-orders by reinforcement learning-based market making.

Lan, Y., Chen, X., & Zhang, W. (2025). News-aware direct reinforcement trading for financial markets.

Leng, J. (2024). AI-driven optimisation of financial quantitative trading algorithms. Applied and Computational Engineering, 100.

Liu, X., Wang, Z., Jin, Y., & Huang, F. (2023). FinGPT: Democratising internet-scale data for financial large language models.

Liu, Y., Chen, H., & Wang, S. (2024). Can ChatGPT overcome behavioural biases in the financial sector? https://doi.org/10.48550/arxiv.2411.13599

Nahar, S., Rahman, M., & Islam, T. (2024). Market efficiency and stability in the era of high-frequency trading. International Journal of Business Management, 1(3), 166.

Seth, A. (n.d.). Artificial intelligence and algorithmic trading: Implications for market dynamics.

Shavandi, H., Ghasemaghaei, M., & Hassanein, K. (2022). A multi-agent deep reinforcement learning framework for algorithmic trading. Expert Systems with Applications, 208, 118124.

Srivastava, A., Kumar, R., & Sharma, P. (2024). AI and algorithmic trading: A study on predictive accuracy and market efficiency. ShodhKosh Journal, 5(1).

Tang, H., Liu, Z., & Chen, Y. (2025). AlphaAgent: LLM-driven alpha mining with regularised exploration to counteract alpha decay. arXiv. https://doi.org/10.48550/arxiv.2502.16789

Vance, R., Thompson, K., & Martinez, L. (n.d.). AI-driven regulatory and risk-control mechanisms for high-frequency financial trading.

Wang, J., Li, M., & Zhang, Q. (2025). Meta-learning reinforcement learning for crypto-return prediction.

Xiao, K., Chen, L., & Wang, Y. (2024). TradingAgents: Multi-agents LLM financial trading framework.

Xiao, K., Liu, H., & Zhang, P. (2025). Trading-R1: Financial trading with LLM reasoning via reinforcement learning.

Zhang, Y. (2025). Interpretable machine learning for macro alpha: A news sentiment case study.

Francis Akpata is the founding Director of Plymouth Capital.