Hedge funds and AI · Hedge funds
Hedge fund performance: how returns are measured and why they flatter
How hedge fund returns are reported, how survivorship, backfill and smoothing biases inflate them, what fees do to alpha, and what the SEC’s Form PF data shows.
On this page
- How a hedge fund's return is calculated
- Survivorship bias: the funds that disappear
- Backfill bias: the history that arrives late
- Smoothing: when prices are estimates
- Alpha, beta and the share fees take
- What the public data shows
- Where these measurements fail
- How Finabase does it
- Frequently asked questions
- Summary
- Check your understanding
A hedge fund's track record is harder to read than a mutual fund's. A mutual fund publishes a price every day and a standardized return in its prospectus. A hedge fund reports privately to its own investors, chooses whether to send its numbers to the commercial databases that build hedge fund indexes, and holds assets that may not have a market price at all. The figures that reach the press are therefore the end of a long chain of choices, and most of those choices make the numbers look better than what investors as a group actually earned.
This lesson explains how hedge fund returns are measured, the three biases that inflate published results (survivorship, backfill and return smoothing), what fees do to a manager's skill, and what the one public, official source on US hedge funds, the SEC's Form PF statistics, does and does not show. Every simulation here uses invented funds; no commercial hedge fund index is quoted. What is a hedge fund? Structure, fees and who can invest covers the structure and the "2 and 20" fee in dollars; this page starts where that one ends.
How a hedge fund's return is calculated
A fund's administrator values the portfolio periodically, often at each month-end, subtracts liabilities and accrued fees, and divides by the units investors hold. The change in that net asset value (NAV) per unit, with nothing added or taken out, is the month's net return. Monthly returns are chained into a year: twelve months of +1% each compound to 12.7%, not 12%.
Three details make a quoted number hard to compare:
- Net or gross. Gross returns are before the management and performance fees; net returns are after them. Under the SEC's marketing rule for registered investment advisers, an advertisement that shows gross performance must show net performance with at least equal prominence, over the same period and with the same method.
- Which investor. Fees differ by share class, side letter and entry date, and a performance fee depends on each investor's own high-water mark. Two investors in the same fund can earn different returns in the same year.
- Which period. The marketing rule requires one-, five- and ten-year returns in advertisements for most portfolios, but it exempts private funds from that requirement. A hedge fund's adviser chooses its periods, subject to the rule's general ban on including, excluding or framing results in a way that is not fair and balanced.
Return is only half of performance. Sharpe ratio: return per unit of risk, calculated on real funds divides the return above the Treasury bill rate by the volatility, and Alpha: the return that market exposure does not explain separates the part of a return that market exposure explains from the part it does not. Both rely on the reported monthly numbers, so both inherit every flaw in them.
Survivorship bias: the funds that disappear
Funds that lose money lose investors, and many close. When a closed fund stops reporting, its history may leave the database, or stay in a "dead funds" file that a casual index ignores. An average taken over the funds alive today is an average of winners (Survivorship bias).
To see how large the effect can be, the chart follows 100 invented funds for ten years. Each fund's yearly return is the Treasury bill rate (3%), plus its exposure to an invented stock market (a beta between 0.3 and 0.7), plus its own alpha, plus noise. The alphas are drawn around zero, so as a group these funds add nothing after fees. One rule decides who survives: a fund closes after any year that leaves it 20% or more below its own peak. The market path is invented too: 12%, 8%, −15%, 20%, 6%, −8%, 15%, 10%, −4% and 9%.
Twenty-four funds closed, eleven of them in the year the market fell 15%. Measured honestly, counting each closed fund's last year and then dropping it, the group turned $10,000 into $14,718, or 3.94% a year. The 76 survivors, averaged over the same ten years, turned it into $16,099, or 4.88% a year. That gap of 0.94 percentage points a year is pure selection: nothing about the survivors' strategy differs, except that their alpha draws happened to be better (0.64% a year on average, against −2.43% for the funds that closed).
The SEC's own data has a related blind spot. Its Private Fund Statistics report notes that Form PF does not require an adviser to tell the SEC when a fund liquidates, so a closed fund may still be counted for up to a year afterwards.
Backfill bias: the history that arrives late
Most databases let a fund add its earlier returns when it starts reporting. A manager has every reason to start reporting after a good run, to attract investors, and little reason to start after a bad one. The good early years are then added, or backfilled, and the index's past improves after the fact. The same choice is sometimes called instant-history bias. It is a cousin of Look-ahead bias: the database's history now includes a selection made later.
The second test starts 50 invented new funds with the same return model. For three years they run privately. Then the 25 with the best three-year record start reporting and backfill those years; the other 25 never appear.
The database's record for years 1 to 3 shows the joiners averaging 6.9% a year, while the full group of new funds averaged 3.1%. In years 4 to 6, reported as they happened, the joiners averaged 3.4%, slightly below the whole group's 3.6%. Their head start was mostly luck. That is the pattern backfill creates in real data: a fund's best numbers are often the ones from before anyone was watching. The same selection, choosing the best of many tries, is why an optimized backtest looks better than its future (Overfitting and backtest bias: 4,094 rules tested on SPY).
Two further choices push the same way. Funds that do well may stop reporting because they are closed to new money and no longer need to advertise, which works against the bias. Funds that collapse can stop reporting before the end, so the worst months may never be recorded at all. How large these effects are in any real database depends on the database and the period, and estimates in the research differ; this lesson does not quote one.
Smoothing: when prices are estimates
A hedge fund that holds hard-to-trade assets, such as private loans, distressed bonds or thinly traded securities, often has to estimate their value. Estimates move slowly. If a month's reported value leans on last month's prices, the reported return becomes an average of recent true returns, and an average moves less than its parts.
A simple calculation shows the effect. Take a fund whose true monthly return averages 0.72% with a volatility of 3.02% a month (about 10.5% a year). If its reported return each month is the average of the last three true months, the reported volatility falls to about 5.8% a year while the average return stays almost the same (0.68% a month). With a Treasury bill rate of 0.25% a month, the Sharpe ratio rises from 0.54 to 0.88 without any change in what the fund owns. The SEC's investor guidance on hedge funds makes the same point from the other side: investors should understand how a fund values hard-to-sell holdings and how much of that valuation is done by independent sources.
Smoothed returns also understate how closely a fund moves with the market, because the market's moves show up in the fund's numbers late and spread out. A low beta and a high Sharpe ratio from a fund with illiquid holdings are both worth checking against how it did in the worst quarter, not the average month.
Alpha, beta and the share fees take
A hedge fund's pitch is often alpha: a return that its market exposure does not explain. The test is the same as for any fund (Alpha vs. beta: market return and the return beyond it): subtract the Treasury bill rate and the return the fund's beta would have earned anyway, and see what is left after fees.
The arithmetic generalizes. A performance fee is charged on the whole gain above the high-water mark, including the part that came from bills and beta, not only on the alpha. So a fund needs a large gross alpha before much of it reaches investors, and a fund with zero gross alpha leaves investors behind the cheap mix by roughly the full fee.
What the public data shows
US hedge funds do not publish returns, but large advisers do report to the SEC. An SEC-registered adviser with at least $150 million in private fund assets must file Form PF, and the SEC's Division of Investment Management publishes the aggregates each quarter as Private Fund Statistics. The edition covering the first quarter of 2026 (filings received through October 9, 2026) counts:
| Measure (first quarter of 2026) | Value |
|---|---|
| Hedge funds reported on Form PF | 10,062 |
| Of which qualifying hedge funds (net assets ≥ $500 million, large advisers) | 2,236 |
| Hedge fund gross assets | $14,659 billion |
| Hedge fund net assets | $6,118 billion |
| Hedge fund borrowing as a share of gross assets | 53.5% |
Gross assets were 2.4 times net assets, a rough measure of how much of the funds' exposure was financed with borrowed money or derivatives. The report also shows how fast money could leave the largest funds compared with how fast they could sell what they own.
At every horizon the funds reported they could raise cash faster than investors could demand it: within 30 days, investors could withdraw 20.7% of net assets while the funds could sell 61.2%. That gap is what lockups, notice periods and gates are for. What the report does not contain is any return series. The SEC publishes the size, borrowing, liquidity and strategy mix of the industry, never how it performed, so any hedge fund performance number in the press comes from a private database with the biases above.
Where these measurements fail
- Every published average is a selection. Funds decide whether to report, when to start and when to stop. No database sees the funds that never reported.
- A short record proves little. With monthly data, a few years of good returns cannot separate skill from luck; Why most active funds underperform: costs, arithmetic and luck shows the same problem with mutual funds, where the dead funds are at least counted.
- Volatility can be manufactured. Smoothed prices lower measured risk and raise the Sharpe ratio; they do not lower the real risk.
- Rare losses hide in short samples. A strategy that earns small gains most months and takes a large loss once a decade can look excellent for nine years.
- Investors earn different returns. Entry dates, fee classes and high-water marks differ, so a fund's published return is not any particular investor's return.
How Finabase does it
Finabase does not show hedge fund returns, NAVs or rankings: hedge funds publish no daily price, and the databases that collect their returns are private and carry the biases described here. What Finabase does show is the public trace of large managers' US stock holdings. The Holders tab on a symbol page, for example Apple's, lists the top institutions and funds holding the stock from their quarterly Form 13F filings, which 13F filings: what institutional holdings reports show and miss explains, including why they are weeks old and show only long positions. A 13F shows what a manager owned, not what its investors earned.
Frequently asked questions
Do hedge funds beat the market?
There is no clean answer from public data. The SEC's Form PF statistics publish no returns, and commercial indexes rely on voluntary reporting that survivorship and backfill push upward. Many hedge funds also aim for lower volatility or low market exposure rather than for beating a stock index, so "beat the market" is often the wrong test. Compare a fund's net return with what its beta would have earned cheaply.
What is backfill bias?
The improvement in a database's history that happens when funds add their past returns as they start reporting. Funds tend to start after strong early years, so the added history is better than what funds of that age earned on average.
Why do hedge funds report returns monthly?
A hedge fund has no daily public price, and the SEC's investor guidance says hedge funds typically let investors redeem four times a year or less. Returns are therefore computed from periodic valuations, often monthly. A monthly figure hides moves within the month and, for illiquid holdings, can lag the market.
Is a high Sharpe ratio proof of skill?
No. A Sharpe ratio from a short record, from smoothed monthly prices or from a strategy with rare large losses can be high without skill. Check how long the record is, how the assets are valued and how the fund did in its worst months.
Where can I find official hedge fund statistics?
The SEC's quarterly Private Fund Statistics report, built from Form PF and Form ADV filings, gives the number of funds, their assets, borrowing, liquidity and strategies. It is aggregated and anonymized and contains no performance data.
Summary
A hedge fund's return is its net asset value change after fees, reported privately and often monthly, and the averages that reach the public are built from funds that chose to report. Losers leave the databases (survivorship), winners arrive with their best years attached (backfill), and estimated prices calm the volatility (smoothing); in invented tests, the first two alone added roughly a point a year and nearly four points to early records. Fees then take most of whatever skill is left. The SEC's Form PF statistics show the industry's size, leverage and liquidity, but not its returns, so any performance claim deserves the same questions: net of which fees, over what period, and counting which funds.
Check your understanding
4 questions on this lesson. Each answer comes with the reason, and nothing is scored.
A database averages the returns of the hedge funds that are still reporting today over the past ten years. Which way is the result likely to be wrong, and why?
- Too high, because funds that closed after losses have dropped out of the average
- Too low, because new funds have not had time to grow
- Neither, because closed funds had average returns
Show the answer
A. Too high, because funds that closed after losses have dropped out of the average Funds that lose money tend to shrink and close, and their records leave the average. In this lesson’s invented test, the 76 survivors averaged 4.88% a year while every fund, counting the 24 that closed, averaged 3.94%. The gap is selection, not skill.
A new fund runs privately for three strong years, then starts reporting to a database and adds those three years. What is this called, and what does it do to the database?
- Backfill bias: it adds a selected, better-than-average early history
- Survivorship bias: it removes the fund’s losing years
- Look-ahead bias: it uses prices the fund could not have known
Show the answer
A. Backfill bias: it adds a selected, better-than-average early history Funds tend to start reporting after a good run, so the history they add is chosen after the fact. In the invented test, the reporters’ backfilled years averaged 6.9% against 3.1% for all new funds, and once live they did no better than the group.
A fund holds hard-to-price loans and reports each month’s value partly from last month’s estimates. What happens to its measured risk?
- Measured volatility falls and the Sharpe ratio rises, though the real risk is unchanged
- Measured volatility rises because estimates are noisy
- Nothing, because the average return is the same
Show the answer
A. Measured volatility falls and the Sharpe ratio rises, though the real risk is unchanged Averaging recent returns makes the reported series move less than the true one. In the lesson’s example, reporting a three-month average cut annual volatility from about 10.5% to 5.8% and raised the Sharpe ratio from 0.54 to 0.88 with the same holdings.
A hedge fund with a beta of 0.4 earns 9.6% before fees in a year when bills pay 4% and stocks beat bills by 4 points. After "2 and 20" it returns about 5.93%. How much alpha did investors receive?
- About 0.33 points, the return above the 5.60% that bills and 0.4 of the market would have earned
- About 4 points, the gross return above the market premium
- About 5.93 points, the whole net return
Show the answer
A. About 0.33 points, the return above the 5.60% that bills and 0.4 of the market would have earned Market exposure alone was worth 4% + 0.4 × 4% = 5.60%, which a cheap mix of index fund and bills could deliver. The fund’s net 5.93% is 0.33 points above that; fees took about 3.67 of the 4 points of gross alpha.
Sources and further reading
- SEC Division of Investment Management, Analytics Office — Private Fund Statistics: Form PF and Form ADV Data, First Calendar Quarter 2026 (Tables 1.1, 1.2, 2.1, 2.3, 2.9, 8.22 and 8.23; filings received through October 9, 2026)
- SEC — Private Fund Statistics (data visualizations page, last updated August 19, 2026)
- SEC — Investment management data (the quarterly Private Fund Statistics report and Excel file)
- eCFR via Cornell LII — 17 CFR 275.206(4)-1, investment adviser marketing (net performance; 1-, 5- and 10-year periods)
- SEC Investor.gov — Hedge funds (fees, redemption limits, valuation)
- Finabase calculations: simulations of invented funds and an invented fee example (no hedge fund index or database is used)
