How It Works For Managers Research Newsletter Apply Now

Seed Capital for Quantitative and Systematic Funds

By Whitespace Research · Published October 2026

Every quant applicant has a good backtest, so seeders diligence the research process instead. Bailey, Borwein, López de Prado and Zhu showed that for a model built on five years of data, a researcher can be misled by looking at "even as few as 45 sample configurations" (Notices of the AMS, 2014). The SEC treats backtested results as hypothetical performance under its Marketing Rule (SEC, 2023). Seeders therefore look past the curve to how it was made: the number of trials, who can change a model, how large the strategy can get, and where the data came from. Until the live record is long enough to speak, the research process is the track record.

A discretionary manager is asked what they believe. A systematic manager is asked how they would know if they were wrong. The questions below are the ones a seeder's research and operational teams tend to work through, roughly in the order they matter.

How many things did you try?

Backtest overfitting is the base rate problem of quant investing. The 2014 paper by Bailey and colleagues showed that searching across model configurations "can end up zeroing in on a model configuration that has a misleadingly good Sharpe ratio," and that the risk appears with far fewer trials than most researchers would guess. Regulators draw the same line. The SEC Marketing Rule defines hypothetical performance to include results "backtested by the application of a strategy to data from prior time periods" when the strategy was not actually used, and requires policies reasonably designed to make such performance relevant to the intended audience.

Keep a research log that records every variant tested, not only the one you kept. Report how much of the data was held out and never touched during development. A seeder who sees a short, honest list of rejected ideas trusts the surviving one more.

Who can change the model?

The Federal Reserve's SR 11-7 guidance, written for banks, defines model risk as "the potential for adverse consequences from decisions based on incorrect or misused model outputs and reports," and asks for "effective challenge" by objective, informed parties (Federal Reserve, April 2011). Large quant firms are not exempt. In January 2025 two Two Sigma advisory firms agreed to a combined $90 million penalty after the SEC found that control weaknesses flagged internally in March 2019, which let staff change model parameters without review, went unaddressed, and that one employee made unapproved changes from November 2021 to August 2023. The firms had already repaid $165 million to affected funds and accounts (Investment Executive, January 2025).

In a three-person firm, effective challenge can be simple: version control on every model, a second person's sign-off before production changes, and a change log the seeder can read. It cannot be absent.

How big can this get?

Capacity is a number you should state and defend. A 2012 working paper by Frazzini, Israel and Moskowitz, using live trades from a large institutional manager, estimated break-even fund sizes in the U.S. of $103 billion, $83 billion and $52 billion for size, value and momentum long/short factors, while short-term reversal strategies did not survive trading costs above about $9 billion (Frazzini, Israel and Moskowitz, 2012 draft). A later paper by the same authors, covering $1.7 trillion of live trades, found actual costs "an order of magnitude smaller than previous studies suggest" (SSRN, 2018).

Two lessons follow. Faster signals have much less room than slower ones, so a high-turnover book needs a lower capacity estimate. And published cost assumptions can be far off, so the estimate should come from your own fills at increasing size, not from a generic cost assumption.

What does the data cost, and where did it come from?

Data is a cost line and a legal exposure at once. In Lowenstein Sandler's 2025 alternative data survey, 90 percent of respondents used alternative data, 81 percent saw cost increases on data products that incorporate AI tools, and possible exposure to material nonpublic information was among the AI-related risks the report listed (Lowenstein Sandler, February 2026). In 2021 the SEC brought its first enforcement action charging an alternative data provider with securities fraud, against App Annie, for using non-aggregated, non-anonymized data to alter estimates it sold to trading firms; the company paid a $10 million penalty (SEC, September 2021).

Put the data budget in the business plan at its renewal price, not its trial price. Keep a file on each vendor: how the data is collected, what consents sit behind it, and what your licence allows. That file is what an operational reviewer will ask for.

What happens when every model says sell?

Systematic strategies share inputs. Khandani and Lo traced the losses of August 2007 to the rapid unwind of one or more large quantitative equity market-neutral portfolios, which spread to other equity portfolios as stop-loss and deleveraging rules triggered (Khandani and Lo, 2007). Know your exposure to the common factors, measure it daily, and show what the book did in the periods when those factors broke.

How quant seed deals are structured

The structure follows the general seed market: a contractual revenue share, a lockup on seed capital of two to three years (Seward & Kissel, July 2026), and working capital support, which appeared in 76 percent of 2025 deals (Opalesque, July 2026). That support matters more in quant than in most strategies, because data, compute and research staff are paid for before the first fee arrives. Seward & Kissel's 2019 study noted that seeders "typically require that a manager reserve investment capacity" in the fund and other products (Seward & Kissel, 2019). In a strategy with a known ceiling, that reservation competes directly with every later allocator, so agree it against your capacity estimate.

When the live record starts to count

The live record starts counting sooner than many quant managers assume. In the AIMA and Marex Emerging Manager Survey 2026, the average track record investors required fell to 1.27 years, and 58 percent would consider a flagship fund with one year of history or less (AIMA and Marex, 2026). A seed that funds the first year of live trading on institutional infrastructure turns a backtest into evidence.

Frequently Asked Questions

Will seed investors accept a backtest instead of a live track record?

A backtest helps explain the strategy, but it is not treated as a track record. The SEC classes backtested results as hypothetical performance under the Marketing Rule. Seeders weigh it alongside the research process, how many variations were tested, how much data was held out, and any live trading period.

What is model risk in a quant fund?

The Federal Reserve's SR 11-7 guidance defines model risk as the potential for adverse consequences from decisions based on incorrect or misused model outputs and reports. In a fund, that includes coding errors, unreviewed parameter changes and models used outside the conditions they were built for. Allocators look for change control and independent review.

How do I estimate the capacity of a systematic strategy?

Start from your own execution data: costs and market impact at increasing trade sizes, by signal speed and market. Research using live trade data found that fast signals such as short-term reversal hit their limits at far smaller sizes than slower factors such as value or momentum. State the estimate and the method.

Why do allocators ask about data vendors?

Data is both a cost and a compliance exposure. In 2021 the SEC brought its first securities fraud case against an alternative data provider, App Annie, over how its data was produced. Allocators want to see how each dataset was sourced and what rights you hold to use it.

Sources

A note on scope: this page describes broad market practice as reported in third-party research and regulatory material. It is general information, not investment, legal or tax advice, and it does not describe terms offered by any specific investor, including Whitespace.

Further reading

Seed capital by strategy

Whitespace reviews systematic strategies from emerging managers. If you are launching one, tell us how the research is run as well as what the backtest shows.

Submit Your Strategy
Published October 2026. Figures come from the publications listed above; survey samples and definitions differ from one source to the next. Check the current edition of each source before relying on a number in an allocator conversation.