This specification describes the 36 benchmarks as published on Oct 7, 2026. Rebuilt every trading night.

How the ETFIQ benchmarks are built

As of Oct 7, 2026, ETFIQ publishes 36 benchmarks in six families, each a written rule over broad, low-cost ETFs. Each family carries a launch date, its freeze date: its tracked record starts with the first close on or after that day, and every figure before it is hypothetical, computed by applying today's rule to past prices. All six families were frozen on October 3, 2026, under rules version 1.3. Third-party model portfolios are what the Model Hub covers. Each model category gets a transparent, rules-based ETFIQ benchmark, used the way the S&P 500 is used for large-cap US stocks. Five family rules and two fund-selection rules generate the whole family. A reader can rebuild every weight from this page, the funds' own filings and their prices.

The hard rules (every benchmark)

Identical for everyone. Rules-based: every weight follows from the written rule and public facts.

Constituents and weights published. No trade files, orders or rebalancing instructions.

No customisation, no fee of any kind, and no payment for inclusion or design.

Never called a recommendation, a model to follow or a portfolio to buy.

Built from broad, liquid ETFs that file N-PORT as registered funds, so the SEC's own monthly returns can check every one. None holds a 1933-Act commodity or crypto trust.

Returns are Tiingo daily total returns, credited "Data sourced by Tiingo" wherever they appear.

How a fund is chosen for a sleeve

Where public data cannot say which funds hold an exposure (a dividend-yield index, a published S&P 500 buy-write index, the MSCI USA factor family), the funds that do, and every fund the doc names or that would win or tie a sleeve if it held the exposure, are listed in data/modelhub/exposures.json, each with its index, the prospectus it is read from, the reviewer's role and the date. That file is an input to the rules, reviewed like any other fact, and both the generator and its independent check read it; no list of funds is written into code.

Each sleeve names an exposure. Two rules choose the funds that fill them, always from the funds ETFIQ publishes (its data at /data/), whose fees, sizes, histories and holdings ETFIQ already reads and checks; a fund ETFIQ does not yet publish is considered once it is published. Net assets are the freshest-dated figure ETFIQ holds for each fund (published at /data/aum.json), each with its date.

What "all sizes" means, as a measurement: the global sleeve's fund must reach the 49.5th percentile of the world market's value in its own book (VT reaches 50.16, ACWI 42.98: large and mid only), and the US sleeve's fund must hold at least as many stocks as make up 99% of the STOCK value (cash and other non-stock lines left out of the total) of the US total-market book ETFIQ uses as its size reference, worked out from that book on every run (1,588 on VTI's book of 2026-08-31, as the independent check counts it on stock value; the selection record prints the count it used). Treasury bills of three months or less: a fund at least 80% in bills with a duration of at most 0.25 years.

Rule A, for every sleeve except thematic:

Among ETFs that hold the exposure, file N-PORT as a registered fund, carry at least $5 billion of net assets and at least 10 calendar years of prices: the lowest expense ratio. When funds tie on the lowest expense ratio, the two largest of them by net assets are held in equal parts.

Why two on a tie: the tied funds cost the same, so splitting between them costs nothing, and it keeps a core sleeve from resting on one issuer's funds, which would measure that issuer's own models against its own products. Two, and the two largest, because the largest are the most widely held way to own the exposure. Net assets are the latest ETFIQ holds for each fund, each with its date, and the generator prints both funds it chose and the funds it left out, with their net assets and dates. Applied to ETFIQ's data on 2026-10-03: US bonds are BND and AGG in halves (AGG, BND, SCHZ and SPAB all charge 0.03%); US stocks are the two largest of VTI, ITOT and SCHB, which the generator prints.

Rule B, for the thematic sleeve: each theme's fund is its largest by net assets that files N-PORT, holds at least $100 million and has at least 3 years of prices. Rule A cannot serve here: most themes have no fund of $5 billion or 10 years, and themes differ by what they hold, not by fee, so the most widely held fund is the theme's fairest representative. Where a theme's two largest funds are within 10% of each other on figures of different dates, the selection record flags that the order could flip at the next re-run. "Within 10%" here and in Rule A's tie flag means the difference divided by the larger figure. On 2026-10-03 two themes are flagged: biotech (IBB $10.5bn against XBI $10.1bn, 4.4% apart) and water (PHO $1.9bn against FIW $1.7bn, 9.3% apart).

The choice is re-run on the last trading day of each December. A change of fund under the same rule is published with its date; it is not a new version of the benchmark. Choosing today's large, cheap, surviving funds and applying them to earlier years is look-ahead, and every hypothetical figure says so.

What the rule picks today, from ETFIQ's data on 2026-10-03 (fees from each fund's 485BPOS, net assets from aum.json):

The six families

Each family is a sleeve recipe at a stock share x. The bands follow the structure of Morningstar's US allocation categories (Morningstar Category for Funds Definitions, April 30, 2025: "strategic equity exposure" of 15 to 30%, 30 to 50%, 50 to 70%, 70 to 85% and more than 85%), never presented as a Morningstar Category. The points x are 20, 40, 60, 80 and 95: each sits inside its band, and they are the equity levels of Morningstar's own Target Risk index family (Morningstar fact sheet, 2021: Conservative 20%, Moderately Conservative 40%, Moderate 60%, Moderately Aggressive 80%, Aggressive 95%), so the convention is the industry's, not ours. Morningstar's 2025 definitions place allocation funds by the volatility they expect, described as similar to those equity exposures, and split US from global funds; ETFIQ keeps the stock share itself, because anyone can check it from a fund's holdings. Each fund's realised volatility is printed beside its band. Every family also has x = 100 where all-stock models exist.

1. Strategic: S(x)

Weights: x% VT, (100 - x)% the US bonds sleeve (BND and AGG in equal parts). All stocks: VT. All US stocks: the US stocks sleeve (the two largest of VTI, ITOT and SCHB, in equal parts). All bonds: the US bonds sleeve.

Rebalance: monthly, on the last trading day, back to the weights.

Why: two funds that ARE the global stock market and the US investment-grade bond market. ETFIQ makes no home-bias call: VT holds the US at its market weight. US bonds, not global, because the models measured are sold to US advisers. Monthly rebalancing keeps the stock share at the band's point, so the benchmark measures the band and not drift.

Measures: strategic and target-risk allocation models (Vanguard Strategic, Core, S&P and Russell series; BlackRock Target Allocation ETF; State Street Strategic Asset Allocation; Invesco Strategic; Schwab Core; Fidelity Target Allocation; Capital Group; JPMorgan Strategic).

1a. Tax-aware: T(x)

Weights: x% VT, (100 - x)% VTEB. Rebalance monthly.

Why: a tax-aware model holds municipal bonds in place of taxable ones, and should be measured against the same choice.

Measures: tax-aware and tax-sensitive allocation models (BlackRock TA Tax-Aware, Vanguard Tax-Aware, Invesco Strategic Tax-Aware, State Street Tax-Sensitive SAA). All-municipal fixed income models: VTEB alone, which belongs to this family and takes its freeze date.

2. Factor: F(x)

Stock sleeve: 20% each of VLUE (value), MTUM (momentum), QUAL (quality), USMV (low volatility) and VB (size), reset to equal weights monthly.

Weights: x% stock sleeve, (100 - x)% the US bonds sleeve. Rebalance monthly.

Why: equal weights across the five factors most widely offered, so no factor is favoured. Version 1 is US stocks only.

Measures: factor and smart-beta models (BlackRock Smart Beta ETF, Fidelity US Equity Factor and Defensive Factor, WisdomTree Multi-Factor Equity, Invesco Dynamic US Factor Rotation).

3. Income: I(x)

Stock sleeve: 50% VYM (dividend stocks), 50% XYLD (option income). Bond sleeve: 50% the US bonds sleeve, 50% SPHY (high yield).

Weights: x% stock sleeve, (100 - x)% bond sleeve. Rebalance monthly.

The yield to beat: the benchmark's own trailing 12-month distribution yield at the same stock share, computed from its funds' distributions and published beside its return. An income model is shown against both: did it pay more, and what did it give up in total return to do so.

Why fixed weights and not weights that move to hit a yield: a rule that chases a yield target buys whatever yields most, which concentrates the risk the benchmark exists to expose.

Measures: income-focused models (Vanguard Income Series, State Street Multi-Asset Income, WisdomTree Global Multi-Asset Income, JPMorgan Income).

4. Tactical: M(x)

Start from S(x). At each month end, each sleeve is tested on its own total-return index (built from the same daily returns the engine uses): if that index is at or above its average over the last 10 month ends, the sleeve is held; otherwise it moves to BIL until a later month end passes the test.

Trades happen at the close of the first trading day after the test, never at the close the test read. Each month's result is one allocation, effective that day, so the benchmark is reset to its weights every month whether or not a sleeve changed. The first test needs 10 month ends of both sleeves, so the tactical history starts at the first month both have one; it does not borrow S(x) before that.

The test is part of the index's own published record. It runs on a fixed calendar, the last trading day of every month, whatever the market has done; its result is published after the test with its inputs (each sleeve's index level, its 10-month average, the resulting weights), and the weights change at the next day's close. It is never a forecast, a view or a call, and it is never published ahead of the index it belongs to.

How it stays inside "no changes timed to market events": nothing in it is decided by anyone when markets move. The calendar is fixed in advance, the one parameter (10 months) was fitted by no one at ETFIQ, and the same arithmetic runs every month.

Why: the simplest widely published trend rule (Mebane Faber, "A Quantitative Approach to Tactical Asset Allocation", 2007, which uses a 10-month average). The parameter was published by someone else before most of the data it is run on; ETFIQ fitted nothing. Faber tested on prices; this uses total return, so a sleeve that pays income is not penalised for it.

Band: a tactical model is placed by its average stock share over its last 36 months (or since its start), because its stock share moves by design.

Measures: tactical and dynamic models (Invesco Dynamic ETF, State Street Quarterly and Active Asset Allocation, JPMorgan Tactical, Richard Bernstein Advisors Macro Allocation, Vanguard Dynamic).

5. Thematic: H(x)

Stock sleeve: equal weights across one fund per ETFIQ theme (the 54 themes on /themes/). Each theme's fund is its largest by net assets that files N-PORT, holds at least $100 million and has at least 3 years of prices. On 2026-10-03, 39 of the 54 themes have one; the other 15 (ai-infra, ai-labs, beauty, coal, cooling, entertainment, health-innovation, humanoid, longevity, memory, millennial, nanotech, pets, photonics, printing) join when a fund qualifies.

Re-chosen on the last trading day of each December, on that day's facts; the new set trades at the close of the next trading day, never at the close it read (as the tactical test does). The whole benchmark, stock sleeve and bond sleeve, is reset to its weights at each quarter end (one rebalance rule per benchmark).

Before the freeze, a theme joins the hypothetical history at the first December after its fund has 3 calendar years of prices, measured on the fund's real trading history (its first trading day), not on a price reader that starts in 2010. The history itself cannot start before the price data the engine reads (2010): the first set is December 2010's, the first December inside that data, traded on 2011-01-03. The early years hold fewer themes and say so.

Weights: x% stock sleeve, (100 - x)% the US bonds sleeve.

Why: equal weights across themes, not across funds or dollars. Weighted by net assets, SMH alone would be 28.2% of the basket ($76.2bn of the 39 picks' $270.8bn on 2026-10-03), GDX the next at 9.6%.

Measures: thematic, megatrend and innovation models.

Which benchmark measures which model

A model is placed by its own look-through stock share, never by its name.

Reference: ETFIQ Strategic at the same stock share

Reference rule, version 1 (2026-10-04). It is a rule for measuring the benchmarks, not for building them: it changes no benchmark's funds, weights, rebalancing or freeze date, and no figure a benchmark publishes about itself.

Every ETFIQ benchmark outside the strategic family is shown against ETFIQ Strategic at the same stock share, so a reader sees what its family's one choice (factors, income, the trend test, themes, municipal bonds) did against the plain mix of the world's stocks and US bonds.

The stock share is the benchmark's own point x: 20, 40, 60, 80 or 95, and 0 or 100 for a single-sleeve or all-stock member.

The reference is the strategic member at the same point: S(x) at 20 to 95, and the US bonds sleeve (All bonds) at 0, so the municipal bonds benchmark is measured against taxable bonds.

At 100 the strategic family has two members, All stocks (VT) and All US stocks (the US stocks sleeve), and the reference is the one the table above assigns from the member's measured look-through US share of its stocks: 90% or more goes to All US stocks, otherwise to All stocks. The share is measured on the member's current allocation (the one in force on the last close its rules read), as the benchmark page's "Stocks by region" rows measure it: for each held fund, its weight times its common stock share of net assets (N-PORT equityPct) times the US share of its stocks (N-PORT usShareOfEquity, the stocks that name a country), summed, over the sum of weight times stock share; a fund with no stocks adds nothing. Every held fund must publish its stock share, and every fund that holds stocks its US share, or the US share is not measured and the member has no reference, with the funds named. The share is tested as printed, to two decimal places, so the printed share and the reference never disagree, and it is printed with the reference and its dates: "against All US stocks, because 96.47% of its stocks are US, as of 2026-06-30 to 2026-07-31". It is never decided from a name.

A strategic member has no reference. It is measured against itself, so its page shows no relative figure; the table below gives the reason.

No neighbour is substituted. Where Strategic has no member at a benchmark's point, or more than one where the rule above does not choose between them, the benchmark has no reference and says why.

The figures, over the same dates: for every window, the difference in total return (the benchmark's minus the reference's, from the same first close to the same last close, in percentage points), with the reference's own return; and, over the window's monthly returns where there are 36 or more, beta, correlation and tracking error (on monthly returns in excess of the 13-week Treasury bill for beta and correlation), and up and down capture. A month or a close the reference does not have makes the figure absent; a window is never cut to fit. They are the same definitions the allocation funds' comparisons use against their band's benchmark.

On the page the reference is named in the row beside the benchmark: "Strategic 60, the reference".

What the rule gives on ETFIQ's data of 2026-10-02:

Stress windows

Stress windows rule, version 1 (2026-10-04). Every benchmark is shown over the market's falls, found by this rule on every run and never chosen by anyone.

The series. The daily total return of VT (the world's stock market) and of AGG (US investment-grade bonds): each fund's adjusted close as Tiingo publishes it (prices with every distribution reinvested), read the way every fund's prices are read here, from the first close on or after 2010-01-04, the first day of the price data the benchmarks are computed from, to the last close.

Peak. A close at or above every earlier close since 2010-01-04 is a peak (the first close is one): the running maximum.

A fall begins when a close is below the latest peak and lasts until the first later close at or above that peak, when the peak is regained. Its peak is the last peak close before it; its trough is its lowest close (the earliest, when two are equal); its depth is the trough divided by the peak, minus one.

Which falls are windows. A fall of VT of 10% or more, and a fall of AGG of 5% or more, measured exactly on the published adjusted closes. A fall of 9.99% in VT is not a window.

The dates. A window runs from its peak close to its trough close, both days included, and is named by them: "Feb 12 to Mar 23, 2020", or "Nov 8, 2021 to Oct 12, 2022" when it crosses a year end (months as Jan, Feb, Mar, Apr, May, Jun, Jul, Aug, Sep, Oct, Nov, Dec; the day without a leading zero), with its fund.

Kept apart. Two falls of one fund cannot overlap or touch: a fall ends when its peak is regained, which is at or before the next fall's peak, so every trough comes before the next peak. A VT window and an AGG window can overlap; both are kept, each with its own fund and dates.

A fall still in progress at the last close is listed once its depth so far meets its threshold, marked "not yet regained", with its lowest close so far as its trough. A later lower close moves its end on the next run, and the day the peak is regained closes it.

What is shown for each benchmark and window: its return from the close of the window's first day to the close of its last, in percent and as what $10,000 at the first close was worth at the last; and its maximum drawdown inside the window, on daily values. A window outside a benchmark's history, or only partly inside it, is not shown for that benchmark ("outside its history"): a window is never cut to fit. Every figure before a family's freeze date is hypothetical.

The windows on ETFIQ's data to 2026-10-02:

The deepest falls the rule leaves out: VT's of 9.67% from Feb 25 to Mar 30, 2026, and AGG's of 4.50% from Jul 8 to Dec 15, 2016.

What ETFIQ will not do

No money from issuers or strategists, for anything: not for coverage, listing, data, design, inclusion or webinars. ETFIQ takes no payment from the issuer of any fund a benchmark holds.

ETFIQ benchmarks are never licensed to issuers as fund indexes or for any other use. A fund may cite a published ETFIQ benchmark as a comparison, as it may cite any published index; ETFIQ is not paid for it and does not tailor anything to it.

Back-tests leave out funds that closed before 2016, because ETFIQ's price source does not carry them. Every hypothetical figure says so, and nothing before 2016 is called free of survivorship bias.

And the red lines (from the SEC's 2022 release IA-6050 and Lowe v. SEC): no trade or rebalance files, no delivery to platforms, no custom versions, no fee tied to assets that follow a benchmark, no licensing to a single fund, and no changes timed to market events.

Governance

ETFIQ publishes its benchmarks the way an index provider publishes an index.

This methodology, with the reason for every choice. Each family carries a launch date (its freeze date), a first value date and a base value of 100, and its back-tested history is labelled hypothetical and kept apart from its tracked history.

A benchmark committee named by role, not by person (the site names no human): the methodology lead proposes, the review lead checks, and the publisher approves any change to a rule.

Changes: a rule is never edited in place. A material change (a new sleeve, a new weight, a new rule) is announced on the methodology page at least one month before it takes effect, and becomes a new version with its own freeze date; the old version's tracked record stays published. A fund replaced under an unchanged rule (the December re-run, a closure) is not a material change and is published with its date.

Annual review of every rule each December, published even when nothing changes.

Errors and restatement: a wrong published figure is corrected and listed on /corrections/ with what changed and why.

Cessation: a benchmark that can no longer be calculated (for example, every fund for a sleeve closes and no replacement qualifies) is ended with one month's notice and its history kept.

Records: every rule version, selection record and generated weight is kept for five years.

A statement against the IOSCO Principles for Financial Benchmarks, self-declared and published with this methodology.

Where to next

Every benchmark

Strategic benchmarks

Tax-aware benchmarks

Factor benchmarks

Income benchmarks

Tactical benchmarks

Thematic benchmarks

ETFIQ Global Stocks benchmark

ETFIQ US Stocks benchmark

ETFIQ US Bonds benchmark

ETFIQ Municipal Bonds benchmark

ETFIQ Factor, 100% stocks benchmark

ETFIQ Thematic, 100% stocks benchmark

Every method

How to check any of this

Every published figure is available as JSON at a stable address on the open data page. Figures ETFIQ computes are free to use with attribution and the as-of date; issuer data stays under its owner's terms. Every number is recomputed nightly from the raw sources by independent code and the result is published on the statistics page. Where ETFIQ has published a figure that turned out to be wrong, it is recorded on the corrections page rather than quietly changed.

ETFIQ is an independent publisher of exchange-traded fund data. It is not a fund issuer, broker-dealer or investment adviser, takes no payment from issuers, and makes no recommendations. A figure marked ETFIQ is one ETFIQ computed; every other figure is the issuer’s or the exchange’s, and is labeled as such on the page it appears on.