How Macro Harbor BEACON Scores Stocks

WHAT BEACON IS (AND ISN'T)

Macro Harbor BEACON is a quantitative stock scanner for the S&P 500 and S&P MidCap 400 (roughly 900 names). It turns fundamentals and price data into a single 0-100 composite score per stock, re-ranked every day.

BEACON is:

• A starting point for your own research
• An educational tool showing how a disciplined sector-relative ranking model behaves across market regimes
• A live track record you can audit on the weekly picks archive

BEACON is not:

• Investment advice, a recommendation, or a signal to trade
• A forecast of future returns
• A substitute for reading filings, talking to management, or understanding the business you're buying

THE COMPOSITE SCORE

Every stock in the universe gets three sub-scores, then a weighted composite on a 0-100 scale.

Sub-score weights inside each bucket

Value = Forward P/E (30%) + EV/EBITDA (25%) + P/FCF (25%) + PEG (20%). Quality = Gross Margin (25%) + ROE (30%) + Debt/Equity (20%) + FCF Yield (25%). Momentum = 6M return (40%) + 52-week high distance (30%) + analyst revisions (30%).

All weights are tagged with a SCORING_VERSION (currently 1.0.0). When we recalibrate, historical scores keep their original version tag so charts stay honest.

SECTOR-RELATIVE RANKING

Raw ratios are meaningless across sectors - a 15x P/E is expensive for a regional bank and cheap for a software company. BEACON ranks every metric within its sector peer group and converts that rank to a 0-100 percentile.

Worked example

Suppose a healthcare stock has a forward P/E of 12. The sector median is 18 and the 10th percentile is 10. We rank 12 against all healthcare P/Es, get the 25th percentile, invert (because lower is better), and return 75. That 75 becomes the Forward-P/E component of its Value sub-score.

Outliers are winsorized at the 5th/95th percentile before ranking. Missing metrics are excluded and the remaining ones re-weighted; if more than half the metrics in a sub-score are missing, the stock is tagged data_quality = partial and excluded from surfaced lists.

Why some sector medians look strange

Because BEACON uses raw medians (not winsorized for the comparison display), some sectors will show negative median values for metrics that intuitively should be positive. The Healthcare sector is the most striking example: its median ROE is currently negative because so much of the sector is loss-making clinical-stage biotech. For a Healthcare stock with a small loss (say, ROE of -1%), the comparison row will correctly read "Better" against the sector median (-7%) because it's less bad than its peers. This is mathematically and economically correct, even though it can look odd at first glance.

For "lower is better" valuation ratios (Forward P/E, EV/EBITDA, P/FCF, Debt/Equity), BEACON does not rank a stock with a negative ratio against its peers. A negative P/E means negative earnings, not "very cheap" — the metric is broken in that range. The stock page shows these as n/m (not meaningful).

THE MEWS REGIME OVERLAY

MacroHarbor's Macro Early Warning System (MEWS) classifies the macro environment into one of five regimes with a confidence score:

• Expansion - growth accelerating, credit expanding
• Late-Cycle - growth peaking, inflation/rates biting
• Stress - risk-off, credit spreads widening
• Contraction - recession underway
• Recovery - turn off the bottom

Each sector has a base multiplier per regime (see the current regime page for the live table). Tech gets a 1.15x boost in Expansion but 0.85x in Stress; Staples get 0.90x in Expansion but 1.20x in Stress.

Confidence tapering

MEWS returns a confidence score (0-100) for its regime call. The effective multiplier is:

effective = 1.0 + (base - 1.0) * (confidence / 100)

So at 100% confidence the full multiplier applies, at 50% confidence half the deviation from 1.0 applies, and at 0% every sector collapses to 1.0 (no regime effect). Borderline regime calls do not cause large rank swings.

RED FLAG GATES

A high composite score does not save a stock from its auditor. Before surfacing a name in the public top-20 or weekly top-5, BEACON applies four gates:

• Going-concern disclosure in the most recent 10-K/10-Q
• Auditor change in the last 90 days
• Covenant violation disclosed in filings
• Earnings blackout - within 5 trading days of a scheduled report, surfaced with a warning but de-prioritized for weekly picks

Red-flagged stocks still appear in the full scanner table (tagged visibly) so you can see them, but they are excluded from headline lists.

HOW WEEKLY PICKS ARE SELECTED

Every Monday morning we lock the Friday-close scores and publish a top-5. The selection rules are:

1. Take the regime-adjusted ranking, highest to lowest
2. Apply a cap of 2 stocks per sector (so one hot sector can't dominate the list)
3. Drop any names flagged by the red flag gates
4. Drop any stock with data_quality = partial
5. Return the top 5 that survive

For transparency we also store the raw top-5 that the ranking would have produced with no regime overlay. Both lists are tracked in the weekly picks archive so you can see whether the regime overlay is actually adding value.

PERFORMANCE TRACKING

Every pick is tagged at publication with the closing price, the SPY price, and the sector ETF price. We then track three horizons:

• 5 trading days - short-term reaction
• 21 trading days - ~1 month
• 63 trading days - ~1 quarter

At each horizon we compute the pick's return and the excess return over (a) SPY, (b) the sector ETF, and (c) the raw-BEACON shadow portfolio. No cherry-picking: every pick is tracked forever, even the losers. The full table is on the weekly picks archive.

DATA SOURCES

• Prices, fundamentals, analyst estimates: Financial Modeling Prep (FMP) - unlimited commercial plan
• Regime classification: MacroHarbor MEWS (internal macro model)
• Universe membership: S&P 500 + S&P MidCap 400, refreshed monthly

LIMITATIONS (IMPORTANT)

• No point-in-time estimate vintages. Our backtests use current analyst estimates, which is a form of lookahead. Forward-looking Momentum sub-scores computed today are NOT comparable to what would have been computed on the historical date in real time.
• Momentum is prospective-only. The Momentum sub-score uses data that was public at the time, but the analyst-revision component is sensitive to the same vintage problem.
• Small sample size. BEACON has been running live since early 2026. Performance numbers are indicative, not statistical.
• US large-cap only. No international, no small-cap, no ADRs.
• No survivorship adjustment in the surfaced universe - delisted names are removed on their delisting date.

AI DISCLOSURE

The narrative text on stock detail pages ("why this score?" and "what the numbers say") is generated by a large language model from the numerical scores and fundamentals above.

The numerical scores, rankings, regime classifications, and performance numbers on this site are not AI-generated - they are deterministic outputs of the scoring pipeline and can be reproduced from the source data.

COMPLIANCE & DISCLAIMERS

Macro Harbor BEACON is an educational research tool. It is not investment advice, not a recommendation to buy or sell any security, and not tailored to your personal financial situation, risk tolerance, or goals. Nothing on this site is a promise, guarantee, or forecast of future results.

Macro Harbor is not a registered investment adviser. You should consult a qualified, licensed professional before acting on any information shown here. Macro Harbor and its principals may hold positions in any security mentioned on this site. Past performance does not guarantee future results.