Quant hedge funds: strategies, the largest firms and the 2007 quant quake

A quant hedge fund is a hedge fund whose trades are chosen by mathematical and statistical models running on computers. People design, test and supervise the models, and the models pick the positions, size them and usually send the orders. The same funds are called quantitative or systematic funds. Renaissance Technologies, D. E. Shaw, Two Sigma, AQR and Capital Fund Management are among the best known.

Pure quant funds managed about 10% of hedge fund assets in 2025 by With Intelligence's count, and the largest multi-manager firms, Citadel, Millennium and Point72 among them, run large systematic teams of their own. A quant fund can hold thousands of positions for a few seconds or for several months. When many funds run similar models they also lose money on the same days, as they did in August 2007 and again in October 2025.

Systematic and discretionary hedge funds compared

Every large hedge fund uses data, and most employ quants. The two types differ in who makes the final decision. In a discretionary fund, analysts research companies or economies and a portfolio manager decides each position, with models as one input among several. In a quant fund the model's output is the trade, and the portfolio manager decides which models get capital and how much risk each may take.

A non-quant hedge fund is a discretionary one: a long/short stock picker who meets company management, an activist, a macro trader who takes views on central banks. Between the two sits a hybrid often called quantamental. A discretionary manager uses systematic screens, risk models and trading algorithms, or a quant fund adds hand-picked data sets and lets people override the model in defined cases. For any fund described as systematic, investors ask how often a person can override the model and under which rules.

Statistical arbitrage, factor models, systematic macro and high-frequency trading

Quant funds run four broad families of strategy, often side by side inside the same firm. They differ mostly in holding period and in the markets they trade.

Statistical arbitrage

Statistical arbitrage bets that prices which normally move together will converge after drifting apart. The simplest form is pairs trading. Find two stocks whose prices have tracked each other, such as two large oil companies, buy the one that has fallen behind, short the one that has run ahead, and close both positions when the gap closes. The technique grew out of work by Gerry Bamberger at Morgan Stanley in the mid-1980s and was run at scale by Nunzio Tartaglia's automated proprietary trading group there. David E. Shaw, a former Columbia computer science professor, joined that group in 1986 and founded D. E. Shaw & Co. two years later.

A widely cited academic test matched US stocks into pairs over 1962 to 1997 and found annualized excess returns of up to 12% for self-financing portfolios of the top pairs, above a conservative estimate of trading costs for most of the period. Later work found the downward trend in profits continuing, though the strategy still did well in long turbulent periods, including the 2008 financial crisis.

Modern statistical arbitrage trades hundreds or thousands of stocks at once. A model strips out each stock's exposure to the market, its sector and common factors, and trades what is left on the bet that unusual moves in that residual reverse within days. One of the simplest versions, the contrarian strategy studied by Andrew Lo and Craig MacKinlay (see our article on random walk theory), buys yesterday's losers and shorts yesterday's winners. A fund running it is supplying liquidity, buying from whoever is selling in a hurry. That is the strategy at the center of the August 2007 quant quake.

Equity factor models

A factor model ranks thousands of stocks every day or week on measurable traits: cheapness against book value or earnings, the past year's return, profitability, how much a company invests, how volatile the stock is. The fund buys the best-ranked stocks and shorts the worst, sized so that market, sector and country exposures roughly cancel. The academic record of these premiums, and how much of it survived publication and trading costs, is in our guide to factor investing.

Quant funds add two layers to the academic version. A risk model estimates how each stock moves with every other, so an optimizer can build a portfolio whose return depends on the chosen signals and little else. And the signals are combined, increasingly with machine learning. In a comparison of methods on US stocks, regression trees and neural networks in some cases doubled the performance of the leading regression-based strategies, because they pick up nonlinear interactions between predictors. Every method leaned on the same dominant signals: variations on momentum, liquidity and volatility.

Leverage multiplies the small edge on each position. Renaissance's institutional equities funds target a long-term average leverage of about 2.5 to 1 and a beta to the S&P 500 of 0.4 or lower.

Systematic macro and trend following

Systematic macro funds trade futures and forwards on stock indexes, government bonds, currencies and commodities. The largest group follows trends, long in markets that have been rising and short in markets that have been falling, with each position scaled to the market's volatility. Man AHL, founded in 1987 by Michael Adam, David Harding and Martin Lueck, began as a commodity trading advisor, and Harding launched Winton in 1997 on the same approach. Other systematic macro models trade carry, holding higher-yielding currencies or bonds against lower-yielding ones, and value across countries. How trend signals work, what the funds earned in 2008 and 2022 and what they charge is covered in our guide to managed futures.

High-frequency trading and market making

High-frequency traders hold positions for seconds or less. Most of the profit comes from earning the bid-ask spread as a market maker, and from tiny, short-lived price gaps between related instruments, such as an ETF and the stocks inside it. Speed decides who gets each trade, so firms rent server space next to exchange matching engines and send data between exchanges over microwave links. When the SEC reviewed US market structure in January 2010, estimates of high-frequency trading typically exceeded 50% of total volume in US-listed stocks.

Most of the largest high-frequency firms, including Jane Street, Hudson River Trading, Citadel Securities and Virtu, trade their own capital and take no outside investors, so they are proprietary trading firms. Their daily results can be very steady: Virtu reported one losing day in 1,238 trading days from 2009 to 2013. When something breaks, the losses pile up in minutes. On August 1, 2012, a faulty order router at the market maker Knight Capital sent more than 4 million orders in the first 45 minutes of trading, and the firm lost more than 460 million dollars.

How a quant fund turns data into trades

Most quant firms split the work into the same five stages.

1. Data. Prices and volumes, company accounts, analyst estimates, and alternative data such as card transactions, satellite images and web traffic. Each series has to be stored as it was known on each past date, so that a backtest never uses a figure revised months later, and it has to include companies that were later delisted or went bankrupt.
2. Signal research. A researcher proposes a pattern, for example that stocks with rising short interest go on to underperform, and tests it on history. Test hundreds of ideas and some will look profitable by chance, the same multiple-testing problem behind academic finance's factor zoo. The usual defenses are a higher statistical bar and a final test on data the researcher has never touched.
3. Portfolio construction. Hundreds of weak signals are combined into one forecast per asset, and an optimizer turns the forecasts into positions under limits on risk, leverage, turnover and concentration. The optimizer needs a correlation matrix for thousands of stocks, and most of a raw estimate is noise, a result physicists established with random matrix theory (see our econophysics overview).
4. Execution. Large orders are cut into small pieces and sent over hours or days to limit their effect on prices. For a fund that turns over its portfolio many times a year, trading costs decide whether a signal that worked on paper earns anything.
5. Monitoring. Live results are compared every day with what the backtest predicted. Signals weaken as other funds find them and trade them away, so the research stage never stops.

A fund whose models make the trading decisions still needs a large research staff. Renaissance has about 300 employees, 90 of them with PhDs in mathematics, physics, computer science and related fields.

The largest quant hedge fund firms

Asset figures for quant firms are hard to compare. The SEC's Form ADV reports regulatory assets under management, a gross figure that counts positions financed with borrowed money, so a leveraged fund reports far more than its investors' capital. Firm websites usually give net figures, and some firms, such as Qube Research & Technologies, publish none. The table gives each firm's own figure where one exists.

D. E. Shaw and Man Group also run discretionary money. Citadel, Millennium and Point72, whose systematic unit is Cubist, each run many quant teams inside a multi-manager structure, and none of the three is a quant fund as a whole. Our list of hedge funds places the quant firms next to the largest macro, multi-strategy and activist managers.

Why quant funds hire physicists and mathematicians

The first quant hedge fund managers came from mathematics. Edward Thorp, a mathematics professor best known for a card-counting system for blackjack, started Convertible Hedge Associates with Jay Regan in November 1969, and the fund was renamed Princeton Newport Partners in 1974. It hedged warrants and convertible bonds against the underlying shares, and Thorp himself called it the first market-neutral derivatives hedge fund. It ran until the end of 1988. Jim Simons broke codes for the US government at the Institute for Defense Analyses and chaired the mathematics department at Stony Brook before starting, in 1978, the firm that became Renaissance Technologies. In 1993 it hired Peter Brown and Robert Mercer, two computer scientists who had worked on speech recognition at IBM Research.

Physicists followed from the mid-1980s. Emanuel Derman, a theoretical particle physicist, went to Wall Street in 1985 and ran quantitative strategies research groups until 2002, a career he described in My Life as a Quant (2004). David Harding studied theoretical physics at Cambridge and Martin Lueck physics at Oxford before they co-founded AHL. In Paris, the physicist Jean-Philippe Bouchaud is co-founder, chairman and chief scientist of Capital Fund Management, which has more than 100 researchers, most of them PhDs, and hires about 15 new PhDs a year, eFinancialCareers reported. The physics these researchers brought, from fat tails to power laws, is explained in our econophysics basics for investors.

The skills carry over directly: estimating models from noisy measurements, deciding whether an effect is statistically real, and writing code that handles very large data sets. A physical system behaves the same way after it has been measured, while a market pattern weakens once enough traders find it and trade it. Pay for these jobs, from new quant researchers to partners, is covered in our hedge fund salary guide.

The August 2007 quant quake

The losses were confined almost entirely to quant equity funds. By the end of the week, Renaissance had told investors that a key fund was down 8.7% for the month, Highbridge Statistical Opportunities was down 18% as of August 8, Tykhe Capital's largest fund had lost about 20%, and Goldman Sachs Global Equity Opportunities had lost more than 30% in a single week. Goldman's fund had a net asset value of about 3.6 billion dollars before the rescue, and the bank blamed "an increase in overlapping trades, a surge in volatility and an increase in correlations". David Viniar, Goldman's chief financial officer, described the moves as 25-standard-deviation events occurring several days in a row. A normal distribution gives even one such day a probability of effectively zero. Real returns have fat tails, and risk models that assume a bell curve understate how often days like these come.

Amir Khandani and Andrew Lo rebuilt the week with the simple contrarian strategy described above, run without leverage. It lost 1.16% on August 7, 2.83% on August 8 and 2.86% on August 9, a three-day loss of 6.85%. Measured against the strategy's 2006 volatility, that was 7.6 standard deviations. On August 10 it gained 5.92%, and the full week ended down only 0.43%. Funds running similar books with several times leverage lost several times as much, and many cut their positions during the fall and missed part of the rebound.

Their explanation: one or more large quant equity market-neutral portfolios were unwound in a hurry, probably a forced sale by a multi-strategy fund or a bank's proprietary trading desk meeting a margin call or cutting risk. Because so many quant funds held similar positions, the selling pushed against all of them at once, and their own risk limits made them sell too. Transaction data later showed that the unwinding began in July 2007 and continued to the end of the year. It started with financial stocks and with portfolios long stocks cheap on book value and short stocks with strong earnings momentum, and market makers temporarily withdrew capital from August 8. Khandani and Lo read the meltdown as that deleveraging and that withdrawal combined.

The value slump and the October 2025 momentum unwind

The value factor, a core signal in many equity quant books, peaked in December 2006 and then lost 57.8% over the next 13 years and 9 months, including 33.7% in the first nine months of 2020.

The most recent episode came in October 2025. Momentum trades that had worked through the year, in AI-linked technology stocks, European banks and gold, reversed as money moved into lower-quality, speculative shares. A Morgan Stanley basket of momentum stocks fell 11.3% in five days, and the selling spread to the quality and low-volatility factors. By October 27, quant long/short funds tracked by Goldman Sachs' prime brokerage desk were down 1.7% for the month, while fundamental equity funds were roughly flat.

Renaissance's funds for outside investors lost far more than that average. The Renaissance Institutional Equities Fund fell 14.39% in October, leaving it down 7.5% for the year after a 22.7% gain in 2024, and the Renaissance Institutional Diversified Alpha Fund lost 15.6%. The equities fund managed just under 20 billion dollars, down from nearly 36 billion at the start of 2020.

Do quant hedge funds beat discretionary funds?

On average they do about as well. From 1996 to 2014, systematic and discretionary hedge fund managers performed about the same once returns were adjusted for volatility and factor exposures. In aggregate, more of the discretionary funds' average return and volatility came from well-known risk factors, the reverse of a common assumption about quants.

The outlier is Renaissance's Medallion Fund, which returned 66% a year before fees and 39% after fees from 1988 to 2021 with a short-term strategy across several asset classes. Medallion closed to outside investors in 2005 and is run for employees. The Renaissance funds that outsiders can buy run long-biased equity strategies, and the larger of them lost 14.39% in October 2025.

How investors get exposure to quant strategies

Quant hedge funds are private funds, open only to accredited investors and, at many large firms, to the wealthier category of qualified purchasers, with minimums that commonly run from 100,000 dollars to 1 million dollars or more. Several systematic approaches are also sold in public funds with daily liquidity and lower fees, usually with less leverage and fewer short positions:

• managed futures ETFs and mutual funds, which run trend-following models (see managed futures)
• factor and smart beta ETFs, which hold long-only versions of value, momentum, quality and low-volatility portfolios (see factor investing)
• mutual funds run by quant managers, AQR's among them

Due diligence on a quant manager covers the usual operational ground in our due diligence questionnaire guide, plus questions specific to models. How much of the track record is live and how much is backtest? How much capital can the strategy absorb before trading costs eat the signal? How much of the return comes from common factors that a cheaper fund also sells? When can a person override the model, and what happened to the fund in August 2007 and in October 2025?

Frequently asked questions

What are the biggest quant hedge funds?

Among firms that run mainly systematic money, AQR reported about 187 billion dollars of assets at the end of 2025, D. E. Shaw more than 115 billion dollars of investment and committed capital in September 2026, Renaissance Technologies 96.4 billion dollars of regulatory assets in its 2026 SEC filing, Two Sigma more than 80 billion dollars and Capital Fund Management more than 36 billion dollars. Man Group, whose AHL unit is one of the oldest systematic managers, runs 270.9 billion dollars across systematic and discretionary funds. The figures use different definitions, so they rank the firms only roughly.

What is the difference between a quant hedge fund and a regular hedge fund?

In a quant fund, computer models choose and size the trades, and people design, test and supervise the models. In a discretionary or non-quant fund, a portfolio manager decides each position after research, and models are one input among several. Both are private funds with performance fees, and most discretionary funds also employ quants for screening, risk and execution.

Do quants work for hedge funds?

Yes, at almost every large hedge fund. At a quant fund, quant researchers build the trading models, software engineers build the data and trading systems and quant traders run execution. At a discretionary fund, quants support portfolio managers with screens, risk models and backtests. Electronic trading firms such as Jane Street and Hudson River Trading hire the same profiles but trade only their own capital.

How much do hedge fund quants make?

Jane Street lists a 300,000 dollar base salary for new quant researchers before bonus, and D. E. Shaw lists salaries of up to 350,000 dollars for open roles in New York. Average pay including bonus in D. E. Shaw's UK entity was over 1 million dollars in the year to March 2025. Pay depends heavily on the firm's results and on how directly a person's work shows up in profit and loss.

Which quant fund has been the most successful?

Renaissance Technologies' Medallion Fund. From 1988 to 2021 it returned 66% a year on average before fees and 39% after fees. It closed to outside investors in 2005 and is run for Renaissance employees. The Renaissance funds open to outside institutions have far lower returns: the Institutional Equities Fund lost 14.39% in October 2025.