Computed fresh after every close.

After each US market close, our data pipeline runs end to end — operator-triggered and human-verified each trading day (typically publishing by early-to-mid afternoon Pacific), with a freshness gate that refuses to compute or publish on stale or incomplete data:

1. Fetch — pull the day's OHLCV bars for approximately 5,500 stored symbols.
2. Verify — check completeness and backfill any genuine misses. A freshness gate refuses to compute on incomplete data, so a bad data day never produces a misleading reading.
3. Compute — recompute every indicator, signal, and dashboard from the verified data (breadth and internals first, since they feed the composite tools like the Hindenburg Omen).
4. Publish — the computed outputs are committed and deployed, so the live site reflects the latest close.

Around that fatal core, a set of non-fatal steps refreshes everything else on its own natural cadence — economic indicators, filings, options snapshots, housing and energy data — so a third-party outage never blocks the market tools. The pipeline also repairs stock splits by re-fetching adjusted history (never by guessing ratios), auto-matches news headlines and fresh SEC filings to the day's volume-scanner flags, and pings search engines with exactly the URLs whose data changed. A separate research step verifies catalysts for the day's biggest movers before they appear on the scanner pages — AI-assisted, human-reviewed, disclosed in how we use AI.

Systematic strategies run only after the daily market data passes the same completeness checks. The workflow keeps the historical simulation and the subscriber experience on the same clock:

1. Close — the model reviews eligible stocks using information available by that session's close.
2. Overnight — Premium members can review the current portfolio, new entries, retained positions, and planned exits before the next session.
3. Next open — the historical record assumes portfolio changes at the following market open. Actual execution can differ.

Performance and high-level Methodology remain public. The complete formula is not: public pages disclose the thesis, validation, execution assumptions, and risks, while proprietary weights, windows, cutoffs, filters, and decision rules stay private. The strategies themselves are no longer published on this site.

Everything is computed from our own database of daily bars for approximately 5,500 stored symbols; approximately 4,750 eligible common stocks enter a typical daily breadth calculation. The database has sixteen years of broad-universe history and SPY back to 1993. Owning the data means we control the universe, the survivorship handling, and the exact calculation — rather than inheriting a vendor's black box.

Source feeds, by pillar:

• Markets & Signals — our own daily OHLCV database (~5,500 stored symbols; ~4,750 eligible common stocks in a typical breadth calculation), which drives the breadth, internals, sector / country / crypto, and intraday session tools, plus FINRA consolidated short interest for the short-interest and squeeze tools.
• Economy — the Federal Reserve's FRED/ALFRED service (with real release vintages), FINRA (margin debt), the OECD (where FRED mirrors froze), the CFTC, Robert Shiller's own CAPE dataset, and the EIA's bulk grid files for the energy pages. Economic series are dated by release, not reference period — charts show prints when the market learned them.
• Fund Managers & Corporate — SEC EDGAR: 13F filings for the fund portfolios, and XBRL filing data for share counts (the buyback tracker) and platform ad revenue.
• Options & Sentiment — OCC cleared-volume data, end-of-day options-chain snapshots, CBOE put/call ratios and volatility indexes.
• Housing & Earnings — Census, NAR (including our reconstructed 1999+ existing-home-sales archive), Zillow Research, Freddie Mac, the Fed's Z.1 accounts, Redfin (attributed estimates), and the Finnhub earnings calendar joined to our own price bars for surprise-vs-reaction reads.

Where a series is a proxy or an estimate — a Redfin model, an ETF stand-in, a reconstructed archive — the page says so in plain sight rather than presenting it as ground truth.

The reference side of the site divides its labor deliberately: Concepts define (what a term means, exactly — formula, worked example, failure modes), Manuals interpret (what a gauge has historically meant, with every famous episode computed rather than retold), and the tools measure (today's reading, recomputed daily). One subject can live on all three layers; each page answers its own layer's question and hands off rather than competing.

Every tool page follows the same shape:

• A dated reading at the top — what the tool says today, in plain English, stamped with the market close it was computed from.
• The chart — the full history behind the reading, interactive.
• The methodology — exactly how the number is calculated, and how to interpret it, so you can check our work and quote it with confidence.

If a served reading ever falls more than one trading day behind, the page shows a staleness notice so you always know which session you're looking at.

Browse the full catalog, dig into our methodology, or check the FAQ.