A forex manager's track record is not one number. Whether the vehicle is a forex managed account or a pooled fund, the record is a bundle of measurements, each answering a narrow question, each silent about everything outside that question. Elsewhere on this site we discuss whether the track record should be the deciding factor at all and offer a quick scan of what a well-organized record looks like. This article does a different job: it goes metric by metric through what a complete record should contain, why each item is there, what it cannot prove, and how the figure is most commonly distorted or quietly left out.
Two framing points before the inventory.
First, regulatory lanes differ. A U.S. commodity trading advisor (CTA) or commodity pool operator (CPO) typically prepares a written performance record; check the firm and its personnel on NFA BASIC and read the firm's own documents [S07]. An SEC-registered investment adviser's advertising is governed by the SEC marketing rule [S12][S13]. A UK retail financial promotion follows FCA COBS 4.6 [S22]. Retail off-exchange forex at a dealer is a separate regime again [S10]. GIPS — the CFA Institute's Global Investment Performance Standards — is a voluntary presentation standard, not a law [S14]. Many legitimate forex managers are not GIPS-compliant; that is not itself a red flag. This article uses these frameworks as reference points for what "complete" looks like, not as claims that any particular manager is subject to all of them.
Second, the past-performance legend is not decoration. Many written performance records include the prominently displayed statement: "PAST PERFORMANCE IS NOT NECESSARILY INDICATIVE OF FUTURE RESULTS." The FCA requires a prominent warning that figures refer to the past and that past performance is not a reliable indicator of future results [S22]. Every metric below should be read with that in mind.
What a track record is for
A track record is an evaluation tool, not a prediction engine. It documents what a program did, under specified conditions, with specified money, net or gross of specified costs. Used well, it lets you test a manager's story against their arithmetic: does the claimed style match the return path, the drawdowns, the exposure? Used badly, it becomes an extrapolation: the last three years, projected forward.
The distinction matters because the measurement itself is noisy. Sharpe (1994) showed that a historic risk-adjusted ratio is tied to the t-statistic of the mean return — short records carry wide error bars by construction [S17]. Lo (2002) showed that the inputs to such ratios are estimated with error, and that in his illustrative hedge-fund example, annualizing a monthly Sharpe ratio the usual way can overstate the annual figure by as much as 65% when returns are serially correlated (per the published abstract) [S18].
So the honest questions a track record can answer are: What happened? Under what conditions? Measured how? Verified by whom? The question it cannot answer is what will happen — which is exactly why that legend is worth taking at face value.
Notably, a complete written performance record is itself a bundle: program identity and start dates, account counts, assets, monthly and annual rates of return, and dated drawdown figures. No framework reviewed for this article treats a single figure as sufficient. If any of these is missing, ask why.
The metric-by-metric inventory
Each family below follows the same pattern: what it is, why it belongs, what it does not prove, and how it gets distorted.
1. Return history and consistency
What it is. The path of periodic returns — monthly, ideally — plus compounded annual and year-to-date figures. Ask the manager for monthly rates of return covering recent calendar years plus year-to-date (or the life of the program if shorter), with annual figures computed on a compounded monthly basis, and ask how the rate of return is defined — often net performance divided by beginning net asset value. For pooled vehicles, ask whether figures are net of fees, expenses, and allocations to the operator. Check how current the figures are, and note the date of the written performance record.
Why it belongs. A single "since inception" compound figure can hide almost anything: one outsized month, a long flat stretch, a change of style. The monthly path is where loss clustering, drift, and dependence on a single episode become visible.
What it does not prove. That the return was earned the way the marketing says, that it is repeatable, or that a new client would have received it. Composite averages are not any individual account's experience.
How it gets distorted. Window selection is the classic move — starting the record at the first profitable month, or quoting the best rolling twelve. The SEC marketing rule prohibits presenting performance results or time periods "in a manner that is not fair and balanced" and prescribes 1-, 5-, and 10-year presentations for most advertisements [S12][S13]. The FCA requires complete 12-month periods covering the preceding five years or the program's life [S22]. Composite construction is the second move: ask whether accounts whose returns differ materially were combined, and GIPS requires composites to include all actual, fee-paying, discretionary accounts that fit the definition — a rule that exists specifically to prevent cherry-picking [S14]. GIPS also prohibits annualizing any period shorter than one year [S14]. A record that violates any of these conventions is not automatically fraudulent, but it is manager-styled, and you should ask why.
2. Drawdowns and recovery
What it is. A drawdown is the losses experienced over a specified period. The worst peak-to-valley drawdown is the greatest cumulative percentage decline in month-end net asset value from a high that is not subsequently equaled or exceeded, expressed with the months and years of the decline. A complete written performance record shows both the largest single-month drawdown and the worst peak-to-valley drawdown for the window covered, each dated. Our separate article explains drawdowns in more depth. Recovery — the time from trough back to the prior peak — is often omitted; it can and should be computed from the monthly table.
Why it belongs. No investor lives at the average. The peak-to-valley figure is the loss you would have carried had you invested at the high, and the dates tell you whether it was a one-month shock or a two-year grind.
What it does not prove. That the worst is behind. It also does not capture intra-month damage: month-end drawdown figures miss a violent intra-month decline that recovered by the statement date.
How it gets distorted. Undated "max DD" figures; drawdowns computed after a strategy change ("since we improved the filter"); and above all, the funding denominator. NFA Compliance Rule 2-34 requires rates of return and drawdowns to be computed on the nominal account size, precisely because a percentage computed on partial cash funding would understate — or restate — the real risk [S07]. A 10% drawdown on nominal size is a 50% loss of cash in an account funded at 20%. Ask for the denominator every time.
3. Volatility and downside risk
What it is. Standard deviation measures the dispersion of returns around their own mean; we cover the standard deviation measurement separately. GIPS requires compliant reports to show the three-year annualized ex-post standard deviation of monthly returns for both the composite and its benchmark, and requires any additional risk measure to use the same period and methodology for both [S14]. Downside deviation measures only shortfalls below a chosen minimum acceptable return (MAR) [S16].
Why it belongs. Two programs with identical compound returns can deliver very different month-to-month experiences. Downside measures ask a different, often more relevant question than total volatility: how badly and how often does the program miss your threshold?
What it does not prove. Low measured volatility does not mean low risk. Serially correlated or infrequently marked returns compress measured volatility and flatter every ratio built on it [S16][S18]. And a quiet sample proves only that the sample was quiet — as Kidd puts it, quoting Sortino, "just because nothing bad happened doesn't mean you didn't take any risk" [S16].
How it gets distorted. Unstated periodicity and windows; volatility computed net for one manager and gross for another (GIPS requires disclosing which, and recommends gross so fee conventions don't masquerade as risk differences) [S14]; downside deviation divided by only the losing observations instead of all observations, which shrinks the denominator and inflates the resulting ratio [S16]; and three-year statistics quoted from records that do not contain thirty-six months (GIPS requires disclosing that shortfall) [S14].
4. Risk-adjusted returns
What it is. Ratios that compress return and risk into one figure. The Sharpe ratio is the average differential return — fund minus a benchmark, classically the risk-free rate — divided by the standard deviation of that differential [S17]. The Sortino ratio replaces the denominator with downside deviation below a MAR [S16]. A Calmar-style ratio (compound annual return divided by maximum drawdown, often over a trailing window) is common industry usage, but its exact definition varies by publisher — before comparing Calmar figures across managers, confirm the window and the formula behind each one.
Why it belongs. A 30% year bought with a 40% drawdown is not the same product as a 12% year with a 6% drawdown. Ratios make that trade-off comparable — when, and only when, the inputs match.
What it does not prove. That the manager fits your portfolio. Sharpe himself noted the ratio ignores correlation with everything else you hold; a lower-ratio strategy with low correlation to your existing book can be the better addition [S17]. Ratios are also silent about tail shape and about anything outside the sample. None of Sharpe, Sortino, or Calmar is a required field under the NFA or SEC rules reviewed for this article.
How it gets distorted. Non-comparable inputs: different risk-free rates or MARs, different windows, gross for one manager and net for another [S14][S16]. Mechanical annualization: multiplying a monthly Sharpe by the square root of twelve assumes uncorrelated returns, and Lo's abstract reports overstatement of as much as 65% in his hedge-fund illustration when that assumption fails [S17][S18]. Smoothed or stale marks inflate every ratio at once. A ratio you cannot rebuild yourself from the monthly net series, with a disclosed benchmark and window, is a marketing number.
5. Trade-level statistics
What it is. Win rate, average win versus average loss, and number of trades. Notably, many written performance records are account-level, not trade-level: they may report how many accounts closed during the window with positive versus negative lifetime compounded returns, plus a variability measure. A dealer-book profitability percentage, when published, describes a retail customer book, not any manager you are evaluating [S11]. Neither is a per-trade win rate.
Why it belongs. A high win rate coexists comfortably with losing money if the average loss dwarfs the average win. Trade-level detail is the only way to see whether the record rests on many independent decisions or a handful of correlated bets — and the size of that sample governs how much the averages mean [S17][S18].
What it does not prove. Skill. Win rate without payoff size proves nothing by itself, and no opened regulatory text requires — or standardizes — per-trade expectancy, so any such figure is computed however the manager chose to compute it.
How it gets distorted. Quoting win rate without the loss side; counting scale-ins as separate winning trades; presenting hundreds of tickets that express one repeated theme as if they were hundreds of independent bets; and platform statements or "live signal" screenshots that cannot be tied to the accounts that make up the official record. Closed-account counts have their own blind spot worth asking about: they describe accounts that closed, while the currently open accounts are still mid-story.
6. Leverage and exposure in a forex managed account
What it is. The relationship between the market risk actually on (notional exposure) and the money that absorbs losses (cash, equity, NAV). A partially funded account is one where the cash deposited is less than the account size that sets the trading level. NFA 2-34 requires performance on nominal size, forbids imputing interest on unfunded notional, and requires CTAs to tell non-institutional clients in plain terms that partial funding increases leverage, raises fees as a percentage of actual cash, and to show the effect with an example or formula [S07].
For context only — this is the retail dealer regime, not a CTA program limit: as fetched 21 August 2026 (rule text listing an amendment effective 18 March 2026), NFA Financial Requirements Section 12 requires forex dealer members to collect a security deposit of at least 2% of notional on listed major currencies and 5% on others, with the higher rate applying to mixed pairs [S10][S11]. Because of leverage, you can rapidly lose all funds deposited and may lose more than you deposit.
Why it belongs. The same strategy run at one-times and at five-times exposure is not the same product. Without the leverage picture, every other percentage in the record is ambiguous.
What it does not prove. A stated leverage figure does not prove the risk is proportionate — exposure concentrated in one pair at 2:1 can be riskier than a diversified book at 4:1 — and a dealer's margin floor says nothing about what a given managed program actually runs.
How it gets distorted. Cash-on-cash returns advertised from thinly funded accounts; "conservative leverage" claims with no notional-to-NAV arithmetic; fee percentages quoted on nominal while the client wired far less; and conflating a retail margin ratio with a program's actual exposure policy.
7. Concentration and correlation
What it is. How much of the profit and risk comes from one currency pair, one bloc (most often the U.S. dollar side of everything), or one theme such as carry or trend — and how the program's returns move with what you already hold.
Why it belongs. Sharpe's original caution applies: standalone ratios ignore correlation, so a program's value to you depends on your existing book [S17]. Concentration answers a related question: is this a portfolio of ideas or one idea wearing many tickets? Even bank regulators build in humility here — the Basel Committee's post-crisis market-risk framework deliberately limits the diversification benefit banks may assume, because measured correlations broke down in stress [S21]. That is a bank-capital rule, not a forex-manager rule, but the lesson transfers: correlations estimated in calm samples understate how positions move together in a crisis.
What it does not prove. A historical correlation is a property of a sample, not a physical constant. "Low correlation to equities" measured over a quiet window proves little about the week you will most need it to be true.
How it gets distorted. Average pairwise correlations quoted without the window or index; "diversified" labels on books whose risk is dominated by a single pair; and the absence of any concentration reporting at all — no rule opened for this article requires a concentration table in a standard written performance record, so if you want pair-level or theme-level risk attribution, you generally have to ask for it as a diligence item, ideally supported by administrator or custodian records rather than the pitch deck.
8. Liquidity and slippage
What it is. The gap between quoted or modeled prices and what a real order of real size would have received — plus the time and cost of exiting in stress.
Why it belongs. Ask whether any results are hypothetical or simulated. Simulated results can under- or over-compensate for market factors such as lack of liquidity; the NFA's hypothetical-results notice adds price slippage explicitly [S09]. On the dealer side, NFA guidance requires slippage practices to be applied symmetrically regardless of market direction and restricts "no slippage" claims to firms that can demonstrate them [S11]. In bank capital, less-liquid risk factors carry longer liquidity horizons and more capital [S21]. Execution quality is, in short, a recognized way for paper results and lived results to diverge.
What it does not prove. Clean historical fills do not prove clean future fills, particularly if assets have grown — which connects directly to capacity, below.
How it gets distorted. Backtests filled at mid or last-print prices; results generated on a demo feed presented as live; slippage and rollover costs omitted from "gross" curves entirely; and execution statistics prepared by the same party whose execution is being measured, with no independent transaction-cost analysis identified.
9. AUM and capacity
What it is. Assets under management, and the level of assets beyond which the strategy's edge is eaten by its own trading costs. A complete written performance record states, as of a stated date, the number of accounts in the program, total assets under the advisor, and assets in the offered program. If those figures are missing, ask why. GIPS counts only actual assets — no advisory-only assets, no uncalled commitments, no double counting, and net of discretionary leverage [S14].
Why it belongs. Percentage returns are size-dependent. Perold and Salomon's analysis of active management (per the published abstract) describes diseconomies of scale: as assets grow, trades grow, execution costs grow, and percentage returns decline; a fixed percentage fee can push a firm to grow past the point that serves existing clients, and new clients dilute old ones [S20]. A beautiful record earned on a small base in thin crosses is evidence about that base, not about the same strategy at many multiples of it.
What it does not prove. Size is not quality in either direction. Large AUM is social proof, not verification; small AUM is not automatically nimble alpha.
How it gets distorted. "Strategy AUM" that quietly aggregates notional overlays, advisory-only relationships, and related programs; peak AUM quoted from a different product; capacity estimates asserted with no methodology; and the most common omission of all — never mentioning that the track record's best years occurred at a fraction of current size.
10. Fees and net versus gross
What it is. The layers between what the trading made and what the client kept. In this article, gross versus net is discussed conceptually — under GIPS, gross-of-fees returns still reflect transaction costs, while net-of-fees returns also deduct management fees, and any model fees used must produce net returns no higher than actual fees would [S14]. Ask whether pool figures are already net of fees, expenses, and allocations to the operator; NFA promotional performance must be net of all commissions, fees, and expenses [S08]; the SEC bars showing gross performance in an advertisement without net presented with at least equal prominence [S12][S13]; the FCA requires disclosure of the effect of charges when gross figures are shown [S22]. For pools, ask for the break-even point: the trading profit needed in year one for a participant to simply get their money back after all fees and expenses.
Why it belongs. Fees are the one certain number in the whole record. They also shape behavior: incentive structures can reward risk-taking with other people's downside, and partial funding raises fees as a percentage of actual cash without changing the dollar amount [S07].
What it does not prove. "Net of fees" is not self-defining. Net of which fees, at which rates, is the question — a composite netted at a founder's fee rate says little about a new client's terms.
How it gets distorted. Gross equity curves with the fee footnote missing; headline management and incentive fees quoted while execution markups, administration, and other expenses sit outside the calculation; and fee-free proprietary results blended with, or shown ahead of, client results — ask whether proprietary results are labeled as such, set out separately after client performance, and accompanied by a discussion of differences in costs, leverage, and trading.
For a practical example, see how introducing-broker payments can affect costs and incentives.
11. Benchmarking
What it is. The standard a record is compared against. GIPS guidance says a valid benchmark should be specified in advance, relevant to the mandate, measurable, unambiguous, and investable, on a total-return basis — and if no appropriate benchmark exists, which is common for unconstrained absolute-return strategies, the honest answer is to say so and explain why [S14][S15].
Why it belongs. "Up 9%" is uninterpretable without context: against what alternative, with what risk? A well-chosen benchmark converts a number into a comparison.
What it does not prove. Beating a benchmark does not prove skill if the benchmark was chosen after the fact, mismatched to the mandate, or measured on a different basis. No commercial CTA or currency index has any official status as the forex benchmark in the sources reviewed here.
How it gets distorted. Peer-group comparisons are the big one. GIPS guidance does not consider peer universes best practice, for listed reasons: self-reported data, survivorship bias as dead funds drop out, mismatched periods, non-investability, and the fact that the "peer median" is a different fund each period — which makes risk statistics computed on it close to meaningless [S15]. Peer comparisons are used anyway across the industry; the difference between honest and dishonest use is whether those limitations are disclosed, and whether the universe is gross or net. Other distortions: price-only currency indexes compared against total-return programs, and benchmarks switched retroactively to flatter the record — a change GIPS guidance says violates the spirit of the standards when done to make performance look better [S15].
12. Tail risk
What it is. Measures of the rare, large loss. Value-at-risk (VaR) is a loss threshold at a stated confidence level; expected shortfall (ES) is the average of the losses beyond that threshold [S21]. These come from bank market-risk regulation — the Basel Committee replaced VaR with ES in its internal-models capital framework precisely because VaR ignored the losses past its own cutoff, creating an incentive to warehouse tail risk; the gap between ES and VaR widens as return distributions grow fat-tailed [S21]. No NFA rule opened for this article requires either statistic on a forex manager's record.
Why it belongs. Currency markets produce jumps that monthly standard deviation barely registers. A record with no tail discussion at all — no stress scenarios, no worst-week figure, no gap-risk narrative — is a record that has chosen not to discuss the thing most likely to end the relationship.
What it does not prove. Any tail statistic is a model output. A historical VaR computed on a calm sample is a statement about that sample.
How it gets distorted. Unstated horizon and confidence level (a one-day 95% VaR and a twenty-day 99% ES are different animals); tail measures computed at yesterday's asset size and yesterday's positions; diversification assumptions across pairs that all become the same dollar bet in a crisis [S21]; and the softest distortion — simply omitting the topic.
13. Track-record length
What it is. How much history exists, and of what kind. A typical written performance record covers the most recent calendar years plus year-to-date, or the life of the program if shorter. GIPS starts at five years (or since inception) and builds to at least ten, and will not compute its three-year risk statistic without thirty-six monthly returns [S14]. The FCA requires five complete twelve-month periods or the whole life [S22]. The SEC's prescribed advertising periods are one, five, and ten years [S12][S13].
Why it belongs. Edge in currency trading is small relative to noise, so sample size does real statistical work: the significance of a mean return grows only with the length and independence of the sample [S17][S18]. Length across different regimes — trending, ranging, crisis — also tests whether the strategy has only ever seen weather it likes.
What it does not prove. There is no bright line at which a record "proves" skill — no such threshold appears in any rule or standard reviewed here, and any claimed one is invented. A long record can still be one regime repeated; a short one is not automatically luck.
How it gets distorted. Annualized figures built from stubs shorter than a year (prohibited under GIPS) [S14]; "since inception" dates that post-date the strategy's actual, less flattering history; predecessor or splice jobs where a record earned elsewhere is presented without the continuity of people and process that, for example, SEC rules demand before predecessor performance may be shown [S12][S13].
14. Independent verification and data integrity
What it is. The question of who stands behind the numbers — as a topic for your diligence, not a claim anyone has verified any particular manager. The spectrum runs roughly: hypothetical or simulated results; extracted sleeves and manager-built composites; proprietary results; unverified manager-prepared records; records prepared under a prescribed presentation regime; and records supported by independent parties — administrators, auditors, GIPS verifiers.
The precise meanings matter. GIPS verification tests whether a firm's policies for composite construction and performance calculation were designed and implemented in compliance, firm-wide; the standard itself states that verification does not provide assurance on the accuracy of any specific performance report — a separate performance examination is needed to test a given composite [S14]. An audit of a fund's financial statements is not an audit of its marketing tearsheet. Ask whether hypothetical results carry a clear label acknowledging hindsight and liquidity limitations; NFA rules go further, generally barring hypothetical results for a system once the member has three months of actual results for it, outside qualified-eligible-person material [S08], and treating hindsight-built multi-advisor composites as hypothetical no matter what "pro forma" label they wear [S09]. Standardized due-diligence questionnaires, such as those published by AIMA, exist to structure exactly these questions early in the process [S23]. Our companion article on the trouble with forex track records looks at the verification problem from the plausibility side.
Why it belongs. Every other family in this inventory is only as good as its data. The first sorting question for any number is: manager-supplied, or independently supported?
What it does not prove. No verification layer is total. "GIPS-verified" does not certify each month's return [S14]; an administrator strikes NAV but does not audit strategy claims; even a written performance record is typically prepared by the firm itself.
How it gets distorted. "Audited" meaning the management company's own financials, not the performance record; "verified" with no verifier named; "pro forma" meaning losers deleted with hindsight [S09]; demo results labeled live; and hypothetical equity curves spliced onto actual history — GIPS-compliant firms must not link actual performance to historical theoretical performance at all [S14].
A short worked example
Everything in this section is hypothetical. The program, all figures, and all dates below are invented for illustration. They are not the results of any actual trading, do not describe any real manager or account, and are not attainable results. Hypothetical figures have inherent limitations: they are constructed with hindsight and do not reflect real execution, liquidity, or fees.
Suppose a manager's record shows: program inception January 2021, monthly returns through June 2026; annual net returns of +14.2%, −3.1%, +22.8%, +6.4%, +9.7%, and +2.1% year-to-date; largest monthly drawdown −6.8% (September 2022); worst peak-to-valley drawdown −14.5% (April 2022 through January 2023); program assets $85 million nominal across 31 accounts; and, among accounts closed during the window, 9 with positive and 4 with negative lifetime returns.
What the inventory does with this. The monthly table lets you date the recovery: if the March 2022 peak was not regained until August 2023, the program spent roughly seventeen months under water — a fact the two headline drawdown numbers never state. The −3.1% year contained most of the −14.5% episode, which tells you the calendar year smoothed over something the peak-to-valley figure caught. The 4-of-13 negative closed accounts sit awkwardly beside a smooth composite line and are worth a direct question: what distinguished those clients — timing, funding level, fees? Then the questions the record cannot answer from inside itself: Is the denominator nominal or cash, and how funded is the typical account? Net of which fee schedule? What was program size in 2021 when the best year was earned — $85 million, or $8 million? And this record begins in 2021, so no part of it shows the program in a violent liquidity event.
A clean-looking hypothetical record, in other words, generates a page of diligence questions. That is the inventory working as intended.
What a track record can't tell you about a forex manager
Suppose every family above checks out: dated drawdowns, nominal denominators, net returns, a sensible benchmark, an administrator, thirty-six-plus months. The record still does not contain:
- The strategy itself. What the manager trades, how decisions are made, and what has changed over time live in the program's written documents and in conversation, not in the return series.
- The people and the operation. Who actually trades, who reconciles, what happens if one person leaves — operational failure is not a return statistic.
- Terms of access. Fees at your size, funding requirements, redemption and notice terms, and the credit and counterparty structure of the account.
- Registration status. A performance table is not a registration check; look the manager up on NFA BASIC or the official register in their jurisdiction.
- Fit. Correlation to your existing holdings, your horizon, your capacity for loss — the record knows nothing about you.
A track record is a necessary exhibit. It is nowhere close to the whole file.
Where this leaves you
Two habits summarize this entire inventory. First, no single metric is sufficient — not Sharpe, not maximum drawdown, not win rate, not length. Each was shown above to be silent about something material, and the academic sources are explicit that even the best ratios omit correlations, higher moments, and estimation error [S16][S17][S18]. Evaluate the bundle, and evaluate whether the bundle's inputs are consistent. Second, sort every number into independently supported or manager-supplied before you weigh it. Most numbers on most tearsheets are manager-supplied. That is not an accusation; it is a filing system — and the burden of moving a number into the first pile belongs to the manager, not to your optimism.
Educational disclaimer
This article is educational material only. It is not investment advice, not a recommendation, not an offer or solicitation of any investment or service, and not legal, tax, or accounting advice. Nothing here is individualized to any reader. ForexFunds.com has not verified, audited, or approved any manager's track record, and nothing in this article implies that the NFA, SEC, FCA, CFA Institute, or any other body approves any manager, metric, or presentation. Regulatory texts referenced were those in effect as fetched on 21 August 2026 and may change. Trading foreign exchange involves substantial risk of loss and is not suitable for all investors. Past performance is not necessarily indicative of future results.
Frequently asked questions
Can I compare two managers using the Sharpe ratios printed on their tearsheets?
Not directly. A Sharpe ratio is only comparable when both figures use the same return basis (net or gross), the same periodicity, the same benchmark or risk-free rate, and the same window — and when neither was annualized from monthly data without accounting for serial correlation, which can overstate the annual figure substantially [S17][S18]. Rebuild both ratios from monthly net returns before comparing.
Does "GIPS-verified" or "audited" mean the returns are correct?
No. GIPS verification tests a firm's policies and their firm-wide implementation; the standard states it does not assure the accuracy of any specific performance report [S14]. An audit of fund financial statements is not an audit of a marketing performance table. Both raise the floor; neither certifies each number. Ask what, exactly, the third party examined.
Is there a minimum number of years that proves a manager is skilled rather than lucky?
No rule or standard reviewed for this article sets one. Presentation windows (recent years plus year-to-date on many written performance records; one, five, and ten years for SEC advertisements) are conventions, not statistical proofs [S12][S13]. Longer, independent samples spanning different market regimes reduce uncertainty [S17][S18]; no length eliminates it.
—
Sources
Bracketed markers above map to the verified source ledger for this draft (source-ledger.md). All sources were opened on 21 August 2026 (America/New_York); S18 and S20 are cited from their published abstracts only. Links below are the exact primary URLs recorded in the ledger.
Note on IDs: S19 is a reserved/uncited ledger ID (an alias of S11), so the gaps in the numbering below are intentional; do not renumber.
- [S07] NFA Compliance Rule 2-34 — CTA performance reporting; nominal account size (NFA)
- [S08] NFA Compliance Rule 2-29 — communications with the public and promotional material (NFA)
- [S09] NFA Interpretive Notice 9025 — hypothetical performance results (NFA)
- [S10] NFA Financial Requirements Section 12 — forex security deposits (NFA; rule text as fetched 21 Aug 2026, listing amendment effective 18 Mar 2026)
- [S11] NFA, Forex Transactions: A Regulatory Guide (NFA)
- [S12] SEC staff, Investment Adviser Marketing — Small Entity Compliance Guide (SEC)
- [S13] SEC, Investment Adviser Marketing, Release IA-5653 (SEC, 2020)
- [S14] CFA Institute, Global Investment Performance Standards (GIPS) for Firms, 2020 edition
- [S15] CFA Institute/GIPS, Guidance Statement on Benchmarks for Firms (rev. July 2023)
- [S16] Deborah Kidd, "The Sortino Ratio: Is Downside Risk the Only Risk that Matters?" (CFA Institute, 2012)
- [S17] William F. Sharpe, "The Sharpe Ratio," Journal of Portfolio Management (Fall 1994; author's reprint, Stanford)
- [S18] Andrew W. Lo, "The Statistics of Sharpe Ratios," Financial Analysts Journal 58(4), 2002 — published abstract only
- [S20] André F. Perold & Robert S. Salomon, "The Right Amount of Assets Under Management," Financial Analysts Journal 47(3), 1991 — published abstract only
- [S21] Basel Committee on Banking Supervision, Explanatory note on the minimum capital requirements for market risk (BIS, January 2019)
- [S22] FCA Handbook, COBS 4.6 — past, simulated past and future performance (as updated 6 Apr 2026)
- [S23] AIMA, Due Diligence Questionnaires (landing page; questionnaire text not reviewed)
