Quantitative vs. Discretionary Investment Management: What the Evidence Says

Quantitative managers make investment decisions through tested, predefined rules; discretionary managers rely on human analysis and judgment, decision by decision. The best direct evidence, a 2017 Journal of Portfolio Management study of hedge fund returns from 1996 through 2014, found the two approaches delivered similar risk-adjusted performance. The broader research record favors rules for consistency: across 136 studies in other fields, mechanical prediction matched or beat expert judgment in the vast majority of comparisons. In practice, most strong investment organizations are hybrids, using models for breadth and discipline and humans for design, oversight, and judgment about when the model's assumptions no longer hold.

What is the actual difference between the two approaches?

A discretionary manager studies companies, markets, and macroeconomics, then forms a view and acts on it. The decision process is flexible and can incorporate anything, but it lives in a person's head, which means it varies with mood, recent outcomes, and conviction.

A quantitative manager encodes the decision process in advance. A model scores investments on measurable characteristics, builds a portfolio within explicit risk limits, and trades on schedule. The process is rigid by design, but it is also consistent, testable against history, and capable of evaluating thousands of securities at once.

The distinction is about how decisions are made, not how much math is used. Many discretionary managers run sophisticated analytics; they remain discretionary because a human makes the final call. Many quantitative models are simple; they are quantitative because the rule, not a person, pulls the trigger.

What does the research show?

Two strands of evidence matter, one from investing and one from decision science generally.

The investing evidence comes from Campbell Harvey, Sandy Rattray, Andrew Sinclair, and Otto Van Hemert, whose study "Man vs. Machine" (Journal of Portfolio Management, 2017) compared systematic and discretionary hedge funds from 1996 through 2014. After adjusting for volatility and exposure to well-known risk factors, the two groups performed similarly. The study also undercut two common beliefs: it found no empirical justification for allocators' documented wariness of systematic funds, and it found that discretionary funds' returns were actually explained more by generic risk factors than systematic funds' returns were. Dispersion across funds was similar in both camps, meaning manager selection matters just as much whichever style you choose.

The decision-science evidence is older and broader. A meta-analysis by Grove and colleagues (Psychological Assessment, 2000) examined 136 studies comparing expert predictions with simple mechanical rules in medicine and psychology. The rules were about 10 percent more accurate on average and were clearly beaten by experts in fewer than 6 percent of studies, regardless of the experts' experience level. Investing is not medicine, but the underlying problem, making repeated judgments under uncertainty with noisy feedback, is the same shape. The finding suggests that whatever insight a human has is usually applied more effectively through a consistent rule than through case-by-case intuition.

A fair reading of both strands: rules do not manufacture skill that is not there, but they apply whatever skill exists without the variance that human judgment adds. Consistency is the measurable advantage.

Where do humans still add value?

The evidence against unaided case-by-case judgment is not evidence against human expertise. People remain essential in at least four places.

First, problem selection and model design. Someone must decide which questions are worth modeling, which data is trustworthy, and which economic logic justifies a strategy. Models do not invent themselves, and bad design assumptions are a human error that no algorithm corrects.

Second, regime judgment. Models are estimated from history, and humans are better at recognizing when something has no historical precedent: a novel policy intervention, a market structure change, a data source that has quietly broken. The judgment call is not "override the model this week because it feels wrong" but "this model's assumptions no longer describe the world, so retire or rebuild it."

Third, risk oversight. Independent humans should monitor what models are doing in aggregate: concentrations, crowding, liquidity, and exposures the model was never asked to consider.

Fourth, everything specific to the client. No model knows that a family has a business sale closing in March, a charitable goal, or a low-basis concentrated position with emotional history. Translating strategies into a particular family's tax, liquidity, and estate situation is human work and always will be.

A hypothetical example: breadth versus depth

Consider a hypothetical comparison. A skilled discretionary analyst covers 40 companies deeply and updates views on perhaps five of them in a given week. A quantitative model scores 3,000 companies every day on the same dozen measurable characteristics, valuation, profitability, price trend, balance-sheet quality, and rebuilds a 200-stock portfolio monthly within risk limits.

Suppose both have the same modest skill: each call is right 53 percent of the time. The analyst expresses that skill in a handful of large positions, so a few bad outcomes dominate results, and the temptation to double down or abandon ship after losses is constant. The model expresses the same 53 percent across thousands of small, independent decisions per year, so the law of large numbers works in its favor and no single error matters much. This illustration is hypothetical and stylized; real-world skill rates are unknowable in advance, and correlated positions reduce the effective number of independent bets. But it captures why quantitative managers prize breadth: small edges compound reliably only when applied many times, consistently.

The discretionary counterpart is equally real: in situations that are rare, idiosyncratic, and information-rich, such as a complex merger or a one-of-a-kind distressed situation, depth can beat breadth, because there is no large sample for a model to learn from.

Why is the practical answer usually a hybrid?

Pure versions of either approach are rarer than marketing suggests. Most credible quantitative firms employ humans to design, supervise, and occasionally retire models. Most credible discretionary firms use quantitative screens, risk systems, and position-sizing rules to contain the damage human bias can do. The Harvey study's finding of similar aggregate performance is consistent with this convergence: each side has absorbed the other's strengths.

For private clients, the hybrid takes a specific form. Rules govern what should be consistent: rebalancing, tax-loss harvesting, position limits, drawdown protocols, security selection within systematic strategies. Judgment governs what is unique: goals, taxes, entity structure, liquidity timing, and the decision of which rules to adopt at all.

At Atlatl Advisers, that division of labor is deliberate. Our Director of Systematic Investments, Senthil Sundaram, Ph.D., was previously Chief Risk Officer overseeing risk for a $60 billion hedge fund portfolio at Two Sigma, and he leads the implementation of the systematic strategies we've developed. Our Director of Investments, Mark Fedenia, Ph.D., Baird Professor of Finance at the Wisconsin School of Business, leads the evidence standards by which strategies are adopted or rejected. The judgment we exercise is concentrated where judgment helps: research evaluation, risk oversight, and each family's plan.

Key numbers

Frequently asked questions

Which performs better, quant or discretionary?The most direct study, Harvey et al. (2017), found similar risk-adjusted performance between systematic and discretionary hedge funds from 1996 to 2014. The case for rules rests on consistency and auditability rather than a proven return advantage.

Are quantitative strategies riskier?Not inherently. The 2017 study found discretionary funds' returns were explained more by common risk factors than systematic funds' returns. Each approach has distinct failure modes: human bias on one side, model error and overfitting on the other.

Can a model handle a crisis it has never seen?Only within the bounds of its design, which is why risk limits and human oversight exist. Well-built systematic processes specify in advance how risk is reduced in stress, while humans judge whether the environment has broken the model's assumptions.

Is stock picking by humans dead?No. Depth can beat breadth in rare, complex, information-rich situations. The evidence mainly cautions against routine, repeated judgment calls made without a disciplined process.

What should I ask a manager about their approach?Ask what is rules-based and what is judgment, how rules are tested, who can override the model and under what protocol, and how live results have compared with backtests. Vague answers to the override question are the most telling.