Atlas Trend US

Atlas Trend US

_{<a href="README.ko.md">한국어</a>}

Owns the funds whose prices have been going up, in proportion to how calm each one is, and sits in Treasury bills the rest of the time.

In one paragraph

Atlas Trend US watches nine ordinary index funds — US shares, foreign shares, emerging markets, government bonds, gold, commodities, an optional property fund, and a place to park cash. Once a month it asks one question of each: has this thing been going up? It keeps the ones that have, in sizes set by how violently each one has been moving lately, and puts everything else in Treasury bills. It does not read the news, it does not have an opinion about any company, and most months it proposes no change at all.

This is the first of the Atlas Trend series. Its siblings run the same philosophy in other markets; they are separate packages with separate track records, because a rule that works on US ETFs has not thereby been shown to work anywhere else.

The methodology

Trend following. The claim being tested is that an asset's own price path carries information about what it will do next — so this manager reads price and nothing else.

Words this page uses, in plain terms:

ETF	a fund you buy like a share. One purchase gets you a whole market
Total return	the price change plus the dividends, so a fund that pays you cash is not scored as if it had lost that cash
Volatility	how much the price jumps around day to day. Not the same thing as risk of loss, but the thing this manager sizes by
200-session average	the average closing price over roughly the last ten months. A price above it is the common shorthand for "in an uptrend"
Rebalance	moving money between holdings to get back to intended proportions

Four rules do the work.

1 — Four horizons vote. For each fund it computes the total return over the last 1, 3, 6 and 12 calendar months. Each one votes: positive, or not. Reading four horizons rather than one is what stops a single unusual month from deciding the year.

2 — Getting in is harder than staying in. A fund the book already holds stays while at least half the votes are positive. A fund the book does not hold gets in only on three of four and a price above its 200-session average. That gap between the two thresholds is where most of this methodology's value is: it is what a sideways market has to cross twice before it can bill you twice.

3 — Size by how much it moves, not by how much it is liked. What survives is weighted by the inverse of its volatility — the calmest holding gets the most money — capped so no single fund dominates. Then the whole risky sleeve is scaled down if its combined volatility exceeds the target, counting the correlations, so equities in three regions are treated as the one trade they largely are. It never scales up.

4 — Cash is an answer. If nothing qualifies, the proposal is 100% Treasury bills. That is a state this system is designed to reach, not a failure to decide.

The nine, and why these nine

One instrument per economic exposure, each large enough that a monthly rebalance is not a market event, and each with enough listed history that a twelve-month signal is a signal rather than an extrapolation.

role	primary	alternates	why this one
US equity	VTI	ITOT, SPY	Total market rather than large-cap only: the trend being measured should be the equity market's, not the index committee's.
Developed ex-US equity	VEA	IEFA, EFA	Separated from US equity because the two decouple for years at a time, and a single global fund would net that away.
Emerging market equity	VWO	IEMG, EEM	The highest-volatility equity leg, which the inverse-volatility weighting will hold least of — correctly.
Intermediate/long Treasuries	IEF	TLT, GOVT	Duration as its own trend, not as ballast. IEF over TLT as the default because 2022 established that long duration is a risk asset with a bond's reputation.
Gold	GLD	IAU, GLDM	The one holding whose behaviour is uncorrelated with both other legs often enough to matter.
Broad commodities	DBC	PDBC, GSG	The inflation leg. DBC for its length of history; PDBC if the investor wants to avoid a K-1.
Listed real estate (optional)	VNQ	SCHH	Off by default: it is largely an equity exposure with rate sensitivity attached, so it adds a name faster than it adds diversification.
Cash	BIL	SGOV, SHV	Where everything not held sits.

The alternates are not preferences. They are used in exactly one situation — the primary is not tradable at the instant being judged, because the corporate action record or the absence of recent bars says so — and the substitution is reported in the decision.

Two overlaps survive this list and are disclosed rather than engineered away: DBC carries a gold weighting of its own, and VNQ is already inside VTI at its market weight before any separate holding of it.

How a run works

One run, one proposal. Nothing below places an order.

flowchart TB
    classDef reads fill:#1e2a44,stroke:#6f9bf0,color:#cfe0ff
    classDef judges fill:#2f2f38,stroke:#9aa0b4,color:#e8eaf2
    classDef proposes fill:#1b4332,stroke:#40916c,color:#d8f3dc
    classDef person fill:#5c4813,stroke:#f6a609,color:#ffe8b0

    WAKE["Aumos wakes it<br/>monthly, or on the review it armed"]:::reads

    subgraph IN["What it reads"]
        direction TB
        BARS["Daily prices for its nine funds<br/>about 14 months back"]:::reads
        CA["The corporate action record<br/>splits and dividends"]:::reads
        BOOK["Your portfolio as it stands"]:::reads
    end

    subgraph JUDGE["What it works out"]
        direction TB
        S1["Are the instruments still themselves?<br/>gaps, splits, an untradable fund"]:::judges
        S2["Four horizons vote per fund<br/>1 · 3 · 6 · 12 months"]:::judges
        S3["Who is in: half the votes to stay,<br/>three of four plus the 200-day line to enter"]:::judges
        S4["Size the survivors by calmness,<br/>cap each one, scale the sleeve to the risk target"]:::judges
        S5["Everything unallocated goes to Treasury bills"]:::judges
    end

    BAND{"Does any holding differ from<br/>its target by more than the<br/>no-trade band?"}:::judges

    WAIT["WAIT<br/>nothing worth trading for"]:::proposes
    REB["REBALANCE<br/>one basket of target weights"]:::proposes

    ARM["Arms its own next review<br/>at the next month-end"]:::proposes

    subgraph HUMAN["Where a person decides"]
        direction TB
        MAND["Aumos judges it against your Mandate"]:::person
        YOU["You approve, or you do not"]:::person
        ORD["Only then does an order exist"]:::person
    end

    WAKE --> IN --> S1 --> S2 --> S3 --> S4 --> S5 --> BAND
    BAND -- no --> WAIT
    BAND -- yes --> REB
    WAIT --> ARM
    REB --> ARM
    ARM --> MAND --> YOU --> ORD

Legend — 🟦 what it reads · ⬜ what it works out on its own · 🟩 what it hands back · 🟧 where a person decides.

Cadence. The monthly rhythm is what this package asks for, not something it can guarantee. Every decision it submits — including the ones concluding there is nothing to do — arms a plan for the next month-end, so it schedules its own next review rather than waiting to be woken. It picks a month-end rather than a rolling thirty days, for the same reason it measures its horizons in calendar months: a rolling interval drifts against month-end, and the drift is invisible. Two things still sit above it. Your Aumos may refuse an arming — that is its own rule and it changes between versions — and when it does, the manager records the refusal and the review happens on your Aumos's schedule. And the interval stored for the installed manager is whatever you confirm on the install screen. If you want this run monthly, check that interval when you install it.

What it needs

Market	US-listed ETFs, in dollars
Broker connection	Alpaca, linked to this fund. The same login relays bars and corporate actions; no separate alpaca data-source document or second copy of its keys is needed
What a key costs	free with any Alpaca account
One setting to get right	feed — use delayed_sip unless you hold a market data subscription, in which case sip. iex is available and is not recommended: it is one exchange's prints, and its daily close is not the closing price of anything
Other settings	the risk target, the per-fund cap, the no-trade band and whether property is included are all yours to set; the horizons and the voting rule are not configurable, because they are the methodology
The book	it reads your portfolio, and the book's own note on whether it is currently all in cash
Your approval	it proposes and never trades. Every rebalance it computes is a proposal your Aumos judges against your Mandate and you approve or refuse

It reads no filings, no news, no analyst estimates, no fund flows and no macro data. The blindness is the design: a package that reached for a headline when the signal was uncomfortable would be a different methodology wearing this one's track record.

What it is bad at

Every methodology has a shape of market it is wrong in. This one has two, and they are opposites.

• Sideways markets. The signal carries no information and the crossings still happen. The system pays the spread on each one and ends the year having traded a great deal in order to arrive where it started. The hysteresis and the no-trade band reduce this; nothing removes it.
• Sharp reversals. A monthly review learns about a crash a month after it starts. It will sell after the fall and buy back after the recovery has begun. This is the premium a trend follower pays for the years in which the same lateness keeps it out of a long decline.
• The quiet one: it is at its most uncomfortable exactly when it is working. The month it moves everything to Treasury bills is a month it will look foolish if the market rebounds, and the discipline of the rule is the only thing standing between the methodology and a discretionary override that would make its record meaningless.

Do not install this if you want a manager that reacts within the week, that will explain a holding in terms of the company behind it, or that you expect to override when its answer is uncomfortable.

Notes

Choosing between the three Atlas Trend packages — what each buys, what each needs, and where they overlap — is set out in the catalogue's manager index (../README.md). Read it before running this one alongside atlas-trend-kr: that package's US-equity role holds a Korea-listed S&P 500 fund, so on one book the exposure arrives twice at a size neither package chose.

Two limits of the data are real and are stated rather than papered over. Premium and discount to NAV are not available from the source this package reads — which is why the universe is restricted to funds where that gap is measured in basis points. And the vendor's dividend adjustment is a claim, so the package checks it against the corporate action record and reports the discrepancy rather than repairing it: a return computed across an unexplained gap is worse than an admitted one.

Aumos shows no returns for this package until it has earned some. A forward track record takes calendar time rather than compute, and nothing on this page is one.