What AZTMM uses, and why
A line-by-line accounting of every external feed that touches the Market Pulse Index, the regime model, and the Daily and Weekly Pulse. Every input is publicly available exchange-grade data or a free-tier API. Nothing on this page is proprietary, and nothing here is redistributed — we read it, score it, and publish derived observations. This page is reconciled against the live published payload; retired inputs are listed, not silently dropped.
01 Why this page exists
Quant research only travels as far as the trust in its inputs. Most retail commentary names a number and moves on. AZTMM publishes the score and the source of every input that built the score, because credibility lives in transparency about data, not eloquence about conclusions. If you can’t see what fed the model, you can’t evaluate the model.
This page lists every external data source we touch, the cadence we touch it on, the failure modes we plan for, and the things we deliberately do not use. If a source ever changes — a feed deprecates, an API rate-limits, a vendor moves — we update the relevant changelog and keep this page current.
02 Macro & rates macro
Consolidated macroeconomic data feeds
Our macroeconomic data feed is the backbone of every macro and rates input on the site. It is free, has no meaningful key cap for our usage tier, and serves the actual government series — not a derived re-publication. We pull the following series on a daily cadence:
- Yield curve — 2-year and 10-year Treasury constant-maturity yields (DGS2, DGS10). The 10y−2y spread feeds the Rates sub-index.
- Credit spreads — ICE BofA US High Yield Index OAS (BAMLH0A0HYM2) and Investment Grade OAS (BAMLC0A0CM). The HY−IG pair is the single cleanest read of corporate-credit risk appetite.
- M2 money supply — the slow-moving liquidity backdrop for the Liquidity sub-index. Published with a lag; treated as a long-cycle trend input, not a near-term measurement.
- US dollar trend — measured via the UUP dollar-index ETF as a published proxy. We disclose this in the feed itself (the payload carries the note “UUP ETF used as USD proxy; do not interpret as DXY level”). Feeds the FX sub-index.
Update window: The macro feed publishes most series at 7–9 AM ET each business day. Our pre-market recompute runs at 09:15 ET, after publish, before the open.
03 Volatility & VIX volatility
End-of-day exchange volatility feeds
Consolidated exchange venues publish a public set of end-of-day CSV files for every index they administer. We use:
- VIX — spot 30-day implied vol on the S&P 500. Drives the Volatility sub-index baseline level (percentile-ranked against its own 1-year history).
- VIX3M — 3-month constant-maturity implied vol. The VIX/VIX3M relationship is our term-structure read: front below far is contango (calm), front above far is inversion (stress).
Update window: Exchange EOD CSVs post after the close and can lag into the evening on some sessions. Our post-close recompute runs at 18:00 ET with a 22:00 ET safety-net pass — deliberately moved off the old 16:30 ET slot in May 2026 after a staleness incident where the EOD bar had not yet rolled at compute time. The VIX value we publish is aligned to the same closed session as the equity closes it sits next to.
04 Sentiment CNN F&G (primary)
CNN Fear & Greed Index
CNN publishes a free daily 0–100 composite that blends seven sentiment indicators (momentum, breadth, put/call, junk demand, VIX level, safe haven, market volatility). It is currently the primary sentiment input to the MPI. We disclose the trade-off plainly: F&G overlaps several dimensions we also measure directly, so its weight in the composite is constrained, and the sub-indicator should be read as “retail mood per a broad public composite,” not an independent survey.
05 Equity & ETF prices prices
Consolidated quote feeds
Our consolidated quote feed is a community-maintained wrapper around public market-data endpoints. It is free, soft rate-limited, and provides:
- End-of-day OHLCV bars for any US-listed equity, ETF, or index.
- Options chain snapshots (strikes, expirations, last, bid, ask, volume, open interest, implied vol) — used by the Options Gravity tracker, not by the MPI.
- Splits, dividends, and corporate-action history for back-adjustment.
Our post-close pull runs in the 18:00 ET window. We retrieve the sector ETF complex (XLK, XLF, XLE, XLV, XLY, XLP, XLI, XLB, XLU, XLRE, XLC) plus broad-market and factor ETFs (SPY, QQQ, IWM, UUP), and feed prices into the Trend, Breadth, Rotation, and FX sub-indices. A design choice we disclose: we do not pull individual S&P 500 constituents — breadth is measured at the sector-ETF level (cyclical-vs-defensive ratios and small-cap relative strength), not from single-stock advance/decline lines. It is a legitimate, documented breadth methodology, but it is coarser than constituent-level breadth, and readers should weigh it accordingly. Soft rate limit: the feed throttles aggressive callers; our pulls run in batches with backoff so we never hit the wall.
06 Institutional options flow + dark-pool prints
Consolidated end-of-session feed
We use a consolidated end-of-session feed for OPRA-routed unusual options activity and institutional dark-pool prints from major venues. This is the only paid layer in the AZTMM stack. It feeds the Daily and Weekly Pulse narratives — it is not an MPI input. The feed delivers:
- Options flow. Every unusual-activity flag, sweep print, and large-premium order routed across OPRA (Options Price Reporting Authority) for the trading session, with greeks, IV, sentiment classification, and venue tag. Delayed 15 minutes from real-time tape.
- Dark-pool prints. TRF-reported (Trade Reporting Facility) block trades that cleared on alternative trading systems, with size, notional, venue, and timestamp. We apply a $1,000,000 minimum notional floor; below that line is retail-sized noise.
We do not name the vendor on public pages except in the per-post Method Note attribution. The reasoning is policy: we do not want vendor branding to read as endorsement, and we do not want subscribers to infer that AZTMM’s edge is “buy this feed.” The edge is what we do with the data — concordance detection, regime-aware scoring, the Conviction Score — not the data itself, which any institutional desk has access to.
07 What we don’t use
- Real-time tape. Every feed we touch is end-of-session or 15-minute delayed. We are publishing research, not running a trading desk.
- Bloomberg Terminal. Not used. The terminal is the gold standard for buy-side research, but its data redistribution license terms make it incompatible with a public publishing model. We get by with the macro feed, exchange volatility feeds, and consolidated quote feeds for everything terminal users would call “public.”
- Reuters / Refinitiv. Not used. Same redistribution issues as Bloomberg.
- Proprietary alpha factors. No vendor-supplied factor models, no off-the-shelf signal libraries, no licensed quant blocks. Every score on the site is computed from the public inputs above using methodology we publish openly.
- Insider data, leaked decks, expert networks, alt-data scrapes. None of that touches AZTMM. Web scrapes of Twitter, Reddit, satellite imagery, credit-card panels, app-download counts — all out.
- SEC EDGAR filings as MPI inputs. Form 4 rollups appear only in clearly-labeled tracker surfaces, never inside the composite. We do not screen 13Fs for trade ideas.
08 Update cadence per source
| Source | Cadence | Window (ET) | Used by |
|---|---|---|---|
| Macro feed (FRED series) | Daily, business days | 7–9 AM publish | Rates, Credit, Liquidity |
| Exchange volatility EOD CSV | Daily, trading days | Post-close (can lag to late evening) | Volatility |
| CNN Fear & Greed | Daily | Continuous | Sentiment (primary) |
| Consolidated quote feed | Daily, trading days | Post-close window | Trend, Breadth, Rotation, FX |
| Institutional flow + dark-pool feed | Daily, trading days | ~4:30 PM (delayed 15 min) | Daily Pulse, Weekly Pulse narratives |
| MPI recompute | Twice daily, trading days | 09:15 & 18:00 (22:00 safety net) | Composite + regime classifier |
09 Failure modes
We plan for sources to fail. Here is what we do when they do.
Macro feed is down or stale
If the macro feed is unreachable or a series hasn’t updated by 10:00 AM ET, the daily ingest carries forward yesterday’s value with a stale=true flag. The Rates and Credit sub-indices are recomputed using the carry-forward, and the Daily Pulse footer notes that the input was stale. If the gap exceeds two business days, we suspend the affected sub-index for the day rather than publish a misleading reading.
Exchange volatility EOD CSV is delayed
The exchange occasionally posts late on partial-day sessions or after disruptions — this is exactly why the post-close recompute sits at 18:00 ET with a 22:00 ET safety-net pass. If the file still hasn’t rolled by the safety net, the composite carries the prior session’s volatility inputs with the stale flag set, and the published asOf stamp stays honest about which session the data reflects. We’d rather publish late and correct than early and stale.
Quote feed throttles or returns gaps
The consolidated quote feed’s public endpoints occasionally throttle aggressive callers. Our ingest uses exponential backoff with retries. If a sector ETF still fails after retries, we fall back to the prior day’s close and tag the affected ticker. If critical tickers fail, we suspend the Trend, Breadth, and Rotation sub-indices for the day rather than compute from a partial universe.
Institutional flow feed lags or skips
If the consolidated flow feed has not delivered by 5:00 PM ET, we wait. The Daily Pulse depends on it directly — without it there is no observations deck to publish. Late delivery delays the Pulse; missed delivery means we publish a partial Pulse covering only the dark-pool side or skip the day with a transparent note. We have never silently published with a missing feed and we will not start.
10 Methodology stability & changelogs
When sources change — a series is renamed, a vendor moves, a new feed is added, an old one is retired — we update the corresponding changelog and this page in lockstep. The changelogs are append-only: every dated entry stays in the record. Stability is achieved by writing down what changed, not by pretending nothing did.
- MPI Changelog — sub-indicator changes, baseline windows, scoring rules.
- Regime Classifier Changelog — regime model state space, training window, feature inputs.
- Pulse Lab Changelog — surface-level changes to the Pulse Lab dashboard.
- Performance Archive — honest record of past calls and outcomes.
11 Legal note
All data used by AZTMM is publicly available exchange-grade or free-tier API data, except for the consolidated end-of-session feed described in section 6, which is licensed for internal analytical use under standard subscriber terms. AZTMM does not redistribute proprietary data. We publish our derived analytics — the MPI, the regime classification, the Conviction Score, the Daily and Weekly Pulse observations — not the underlying tape, not the raw flow records, not the institutional prints themselves.
Aggregate statistics, anonymized concordance counts, and named-ticker observations consistent with normal market commentary are within fair-use research conventions. We do not publish complete record dumps, programmatic exports, or any output that would let a reader reconstruct the source feed. If a vendor licensing change affects what we can publish, we will update this page and the relevant changelog before the next Pulse goes out.
AZTMM publishes market commentary and research for educational purposes under the publisher exemption established in Lowe v. SEC, 472 U.S. 181 (1985). We are not registered investment advisers. Nothing on this site is a recommendation, solicitation, or offer to buy or sell any security or derivative.