How To Guide

Most screeners answer "what looks good today?" by ranking stocks on current metrics. We split that into two slower, more honest questions:

1. Quality Universe — Which companies have proven they create value consistently for at least a decade? Refreshed monthly. Currently 145 stocks.
2. Today's Buy Zone — Of those quality companies, which are currently available at a reasonable price? Refreshed every weekday after market close.

Quality changes slowly. Price changes daily. Two cadences, one workflow.

Most screeners apply the same thresholds to every stock. We classify each business into one of five archetypes first, then score it on the metrics that actually matter for that kind of business. A toll road shouldn't be judged on growth rate; a cyclical shouldn't be judged on current-quarter margins.

Some companies don't fit any archetype cleanly — those are excluded with a label of uncategorized rather than force-classified. "Uncategorized" is a feature, not a bug.

The number you see in the Universe table (e.g. 82) is a composite of 5 archetype-specific dimensions, each scored against the metrics that matter for that business type. See it broken down on any ticker detail page.

A score in the 80s for a Toll Road means something different than 80 for a Cyclical — same number, different criteria. Compare scores within an archetype, not across.

We don't compare a stock's P/E to the S&P average, or to its industry peers. We compare it to its own 5-year history. A P/E of 20x is the 90th percentile for one stock and the 10th percentile for another. Each stock is judged against itself.

Cyclical override

Cyclicals use P/Normalized Earnings (not reported earnings) for the percentile, then add a guard: if the stock's operating margin is at or above its 10-year average, the grade is downgraded one step and flagged om_at_peak. This is the anti-trap: a cheap P/E on peak earnings is exactly when cyclicals look most tempting and most dangerous.

For every stock we store 60 monthly snapshots of P/E (or dividend yield, or P/Normalized E). When the daily gate runs:

1. Today's fresh price is divided by the cached TTM earnings to get today's metric.
2. We compute where today's metric ranks within those 60 historical values: 0 = at the bottom of the 5yr range; 100 = at the top.
3. That rank is the "percentile" you see in the UI.

Everywhere the percentile appears in the UI, the better side is labelled in green (cheap or rich yield) so you don't have to remember the direction.

When a stock first enters Deep, In, or Approaching, we snapshot its price, P/E (or yield), and the date. Until the grade changes again, those values are the entry.

1. Start in Today's Buy Zone when you have time and capital to deploy. The Deep section is the smallest, most actionable list.
2. For each name that catches your eye, click into the Ticker Detail page. Verify the 5yr P/E chart shows today's value at a real extreme (not a sliver above the band). Check the score breakdown — is any dimension scoring weakly?
3. Read the Zone history at the bottom of the detail page. Is this the stock's first dip into Deep in years, or is it a regular visitor?
4. Use Quality Universe for browsing — filter by archetype if you have a preference, or by zone if you want to scan all currently-attractive names regardless of archetype.
5. The screener doesn't make a buy decision for you. It produces a short list with calibrated valuation context. The investment thesis still happens in your head (or in the deeper research app).

• Qualitative risks — moats erode, AI disrupts incumbents, regulators move. None of that shows up in numbers until it's already in the financials.
• Archetype migration — companies change as they mature. A reinvestment-growth machine eventually becomes a mature compounder, then a cash cow. The classifier flags this with archetype_migration when it sees it, but the boundary is judgment-laden.
• The universe is monthly — a quality-deteriorating stock can stay on the list for up to ~30 days after its fundamentals turn. By design: we screen for durability, not for snapshots. But it's a real lag.
• Holding companies and financials excluded — banks, insurance, REITs, and conglomerates like BRK don't fit the framework. Payment networks (V, MA), exchanges (ICE, CME), and credit ratings (MCO) are kept — they're toll roads, not banks.