Quantitative strategies and the capacity question

Every quantitative pitch deck carries a capacity number, and very few explain where it came from. With institutional demand for systematic strategies building — trade coverage this spring describes managers responding with hard and soft closes, raised minimums, and tightened redemption terms — the spread between honest and marketed quant fund capacity has become one of the more consequential numbers in allocator diligence, and one of the least examined.

Capacity is not a marketing adjective. It is an engineering estimate with known inputs — and, for a registered adviser, a stated number is a factual claim regulators can demand be substantiated.

What an honest capacity number is made of

The academic literature gives capacity a precise anatomy. Bonelli, Landier, Simon, and Thesmar derive it in closed form: because transaction costs scale non-linearly with trade size, a strategy’s realized Sharpe ratio decays as assets grow, and the decay is slower when the strategy trades liquid names, runs on slow-fading signals, and starts from strong frictionless performance. For an investor willing to accept a 30 percent reduction in Sharpe, deployable scale is proportional to the frictionless Sharpe divided by price impact and the speed at which the signal fades. One implication is uncomfortable for marketers of fast strategies: the authors find a slow-burning quality signal supports capacity an order of magnitude larger than value- or momentum-type signals, purely because the signal persists.

These inputs are measurable, not theoretical. Frazzini, Israel, and Moskowitz, working from $1.7 trillion of live trade execution data across 21 developed equity markets, found real-world trading costs an order of magnitude smaller than earlier academic estimates — and found that portfolio construction designed around costs raises net returns and capacity substantially without material style drift. The finding cuts both ways: capacity is larger than older estimates implied, and it is genuinely estimable. A manager who cannot show the work has either not done it or prefers not to share it.

The third input sits outside any one firm’s control. Khandani and Lo’s study of the August 2007 quant unwind documented how crowding in similar long-short equity strategies turned a rapid unwind into a sector-wide cascade. Effective capacity is a property of everyone running the trade. A fund can sit comfortably inside its own modeled capacity and still be outside the strategy’s.

A marketed capacity number is a statement of fact

For SEC-registered advisers, the Marketing Rule — Advisers Act Rule 206(4)-1, adopted in December 2020 — prohibits including in an advertisement any material statement of fact the adviser lacks a reasonable basis for believing it can substantiate upon the Commission’s demand. A capacity figure shown to prospects is a material statement of fact. The rule also defines hypothetical performance to include model results, backtests, and targets or projections — the raw material most capacity figures are built from — with audience-tailoring and policies-and-procedures conditions attached.

Enforcement here is not hypothetical. In September 2023, the SEC charged nine advisers in a single sweep for advertising hypothetical performance to mass audiences without the required policies and procedures, collecting $850,000 in combined penalties. An April 2024 examinations risk alert catalogued further deficiencies, including advisers inaccurately reporting on Form ADV whether their advertisements contained hypothetical performance.

The futures wing of the systematic world answers to a parallel regime: CFTC Regulation 4.41 and NFA Compliance Rule 2-29 require registered CPOs and CTAs to place a prescribed, prominent disclaimer — results designed with the benefit of hindsight, not actual trading — immediately alongside any simulated or hypothetical results.

Two caveats keep the analysis honest. These frameworks are jurisdiction-dependent: they reach SEC-registered advisers and registered CPOs and CTAs, not every offshore manager. And none of these regimes requires a manager to compute or disclose capacity at all. That asymmetry is the heart of the problem. Disclosure is voluntary — but once a number is in the deck, the substantiation burden and the antifraud provisions attach to it.

When capacity is scarce, watch who keeps it

The documented failure mode is not an overstated number. It is quiet reallocation. In December 2020, BlueCrest Capital Management agreed to pay $170 million — all of it returned to investors — to settle SEC charges that, over more than four years, it moved most of its highest-performing traders out of its flagship external fund and into an internal proprietary fund trading personnel capital, replacing their capacity in the external fund with an undisclosed replication algorithm that produced significantly less profit at greater volatility. As the co-chief of the SEC’s Asset Management Unit put it: “BlueCrest investors were marketed a fund with exceptional trading talent but instead got a fund with an undisclosed algorithm that performed worse than those touted traders.” When an edge has limited capacity, the strongest economic incentive in the building is to keep it for insiders.

The substrate beneath the claim matters too. In January 2025, Two Sigma entities agreed to a $90 million penalty — after voluntarily repaying $165 million to affected funds and accounts — over model vulnerabilities employees had identified by March 2019 that went unaddressed until August 2023, during which one employee made unauthorized changes to more than a dozen models. Every capacity claim rests on models behaving as described; technology and change-control diligence is a subject of its own, but a capacity number generated by ungoverned models is substantiated by nothing.

The questions that separate honest from marketed

Allocator-side practice is converging on quantified evidence. One due diligence framework published this February tells allocators to demand from statistical-arbitrage managers a capacity map — expected slippage by AUM level, turnover, and liquidity bucket — on the logic that a strategy “scales until it doesn’t” — and separately cautions that even trend-following’s liquidity, better than credit’s, is not infinite. That is the right form of question: strategy-specific, quantitative, falsifiable.

In practice, five asks do most of the work:

• The capacity model itself — impact assumptions, participation limits, and signal-decay estimates, not the output number alone.
• Realized slippage against modeled slippage at current assets, by liquidity bucket.
• The stated policy as assets approach the limit — soft close, hard close, fee changes — and, specifically, who absorbs any capacity that frees up: external investors or internal capital.
• Model change-control evidence: who can alter production models, and under what review.
• Crowding monitoring: how the firm measures the footprint of neighbors running the same trade.

The takeaway

Honest capacity arrives as a dated estimate with assumptions attached; marketed capacity arrives as a round number with adjectives. The first survives the question “show us realized slippage by AUM” — the second changes the subject. With trade coverage this spring citing more than a quarter of institutional investors planning to increase quant equity exposure, the discipline to ask is worth more than usual, because scarce capacity is precisely the condition under which the BlueCrest incentive operates.

SetOne Labs pressure-tests capacity claims and the diligence posture around systematic strategies, for allocators and for the managers preparing to face them. To examine one in confidence, begin a conversation.

SetOne Labs provides advisory services and general information. Nothing here is legal, tax, or investment advice.