Commodity cycle

commodity-cycle

macro-basket paints the whole cross-asset tape. commodity-cycle zooms in on a single commodity and answers one question: is it in a winning or losing macro setup right now, and which driver dominates it.

You give it a commodity ETF (default GLD) and it pulls the macro context that commodity actually responds to: the dollar (UUP), a real-yield proxy (TIP minus IEF), and for gold the miner and silver co-move set (GDX, SLV). It reads the rolling correlations of the commodity to each driver, folds in its own momentum, and states a one-line take: constructive, neutral, or headwind, naming the dominant macro variable.

This is descriptive, not a forecast. It grounds the read in real ETF prices so an LLM does not have to guess whether the dollar or real yields are the thing hurting gold. It would have flagged a gold drawdown two weeks early: a strengthening dollar plus rising real yields against a commodity that is inversely correlated to both is a headwind before price confirms.

When to invoke

• The session question is about one commodity specifically: "is gold in trouble", "what's driving crude", "is the dollar the problem for gold", "is silver set up well"
• You want the dominant macro driver named, not just a basket ranking
• Confirming or explaining a commodity move: "gold is down, is it the dollar or real yields"
• The user says "commodity setup", "gold cycle", "is this a headwind for gold", "why is gold lagging its miners"

For the full cross-asset dashboard (rates, credit, the dollar, gold, and broad commodities ranked together), use macro-basket. commodity-cycle is the single-name drill-down; macro-basket is the whole tape. For the equity side use market-regime.

What you need

• Nothing required beyond a key. Defaults cover the standard gold run.
• MASSIVE_API_KEY exported in the environment.
• Any stocks tier (all instruments are US-listed ETFs). On Free Basic pass --sleep 13 so the pull stays under the 5-calls/min cap.

Optional:

• --ticker (default GLD): target commodity ETF (GLD/SLV/USO/DBC or any)
• --window (default 60): lookback in trading days for returns and rolling correlations
• --sleep (default 0): seconds between calls for Free Basic

What you get back

Two output layers from one run.

Layer 1: canonical JSON matching output-schema.json. A signals block (dxy_correlation, real_yield_correlation, momentum_quintile, plus miner_divergence and silver_comovement for gold or broad_commodity_comovement for others), a setup label, and the composed take. UIs and downstream agents consume this.

Layer 2: rendered note: the setup label, a drivers block, then the take. See references/rendering.md.

How it works

1. Pull daily aggregates for the target commodity plus UUP, TIP, IEF, and the co-move partners (GDX and SLV for gold, DBC for others) over max(window, 252) * 1.6 calendar days, via /v2/aggs/ticker/{T}/range/1/day/{from}/{to}?adjusted=true.
2. Rolling correlations over the window: commodity vs UUP (the dollar), and commodity vs the TIP-minus-IEF daily return spread (real yields).
3. Gold-specific reads (skipped gracefully otherwise): GLD vs GDX relative return (miner divergence) and rolling GLD vs SLV correlation (silver co-movement).
4. Momentum quintile: the commodity's own window return ranked into a quintile 1..5 against its trailing-year window returns.
5. Compose the take: score each driver by its directional effect (correlation times the driver's own move) plus momentum and miner confirmation into constructive / neutral / headwind, naming the dominant variable. Methodology in references/methodology.md.

Foundations used

• massive-api-patterns for REST auth, rate-limit handling, and the /v2/aggs daily endpoint conventions.

Output mode: note

The deliverable is a single-commodity read with a headline setup and a short drivers block. A note (setup label, drivers, take) fits it better than a wide table; the ranking view is macro-basket's job.

Endpoints used

• GET /v2/aggs/ticker/{ticker}/range/1/day/{from}/{to}?adjusted=true Daily closes per instrument. One call for the target plus the macro context and co-move set.

Doesn't handle (yet)

• Cash-market rates. The real-yield read is the TIP-minus-IEF ETF return spread, not the actual real yield in basis points. Directionally right, not a cash-market substitute.
• FX beyond the dollar index. UUP covers the broad dollar; no per-pair FX. Queued.
• Non-gold miner/co-move sets. Only gold gets the miner (GDX) and silver co-move reads; other commodities fall back to a broad-commodity (DBC) correlation. Oil-services or ags co-move sets are queued.

These are clean PR extensions. The output schema is forward-compatible.