analyze-stock-kpi · KPI demo dashboard

Why these charts?

Why these charts?

CNN F&G rolling (TTM)

Trimmed to a strict 12-month trailing window before render, so the span stays consistent regardless of when the dashboard is opened.

Long-term context (merged)

One chart for multi-year regime context. On the 0–100 left axis: the CNN F&G monthly median (history backfilled via the whit3rabbit/fear-greed-data mirror) and the 5s10s slope (10y − 5y curve-steepness from yfinance ^TNX/^FVX, normalized onto the same 0–100 scale). On a logarithmic right axis: a broad-equity index — the SPY ETF's return rebased to 100, a derived series (never the S&P 500 index level). All three are aggregated to a shared monthly grid.

Backtested long/short 25/25 (hypothetical) — two series

Two point-in-time backtested long/short books — long the top 25 / short the bottom 25, equal-weighted 1/25 per leg (gross 200 %, net 0), marked daily from closing prices, costed at 10 bp one-way turnover, published gross and net. Series A (the headline) ranks by the live qte77 Score at each genuine data-branch snapshot date — the same score shown in the Score column of "Browse all stocks" — trading the first close strictly after the snapshot; no reduced-input approximation. "Latest picks" at the top is a separate, non-historical display: the live aggregated-scores-best / -worst lists (same data as "Browse all stocks", never a second independent fetch), so today's candidates always carry the identical qte77 Score as their source universe (ADR-0014). Series B ("Reconstructed backfill", under "How it's tested") extends the book back to ~2023 on a weekly rank grid across all base universes, ranked by a reduced point-in-time score (score_bt, ADR-0013) computed only from what was knowable on each historical rank date — 6 of 9 Score inputs; forward P/E, PEG and beta have no free point-in-time history, so it is not the live qte77 Score. The two series are never spliced or drawn on one axis — Series B is a labelled approximation, not a continuation of Series A. Each chart marks its primary cadence's rebalance days with a triangle; the collapsible "Rebalance log" under "How it's tested" lists every rebalance newest-first — trade date, reason (initial / scheduled / quarterly-after-filings / buffer exit / buy & hold), and which tickers entered or exited the long and short legs, plus that day's turnover — so it's always clear when a rebalance happened, why, and on which side. Both series are hypothetical and not investment advice: survivorship and a "today's index membership" size bias are disclosed in the caveats under each chart. Both extend forward every Saturday under identical rules.

Why these universes?

Static lists (sp500, eurostoxx, japan, south-america, south-korea, qte77-watchlist)

Hand-curated regional / sector ticker lists. The weekly snapshot fetches fundamentals for every member; "empty" means the cron hasn't run yet.

federal-contractors

Orchestrator-driven: chains usaspending.gov → SEC EDGAR → yfinance to assemble the DoD Top-25 + curated seed. Rebuilt weekly.

aggregated-scores-best / -worst

Cross-universe ranking by the qte77 Score (screener_score) — the same score shown in the Score column and detail panel on every universe. Top 25 + bottom 25 of the merged ticker pool, using today's fundamentals only — the same per-ticker records and score as their source universe's own snapshot for the same date (ADR-0014), never a second independent fetch. This is also "Latest picks" at the top of the page; Series A's chart/metrics in the backtest section reuse this exact ranking construction historically, at each genuine snapshot date. Series B (the reconstructed backfill) ranks all base universes independently, with its own reduced point-in-time score (score_bt) — see Why these charts?.

Backtest rules

Series A (genuine decisions). Hypothetical, point-in-time backtest — not investment advice. Series A's chart and metrics simulate long the top 25 / short the bottom 25, ranked by the live qte77 Score at each genuine data-branch snapshot date (the same construction as the aggregated-scores-best/-worst universes), traded at the close of the first trading day strictly after the snapshot. Equal-weighted 1/25 per leg (gross 200 %, net 0), marked daily from closing prices, costed at 10 bp one-way turnover, published gross and net. The book is held across a snapshot gap. Returns are total returns (dividends included) in local currency; survivorship and a "today's index membership" size bias are disclosed in each series' caveats under "How it's tested". Extended forward every Saturday. "Latest picks" at the top is a separate, non-historical display: it shows the live aggregated-scores-best / -worst lists (full qte77 Score, same data as "Browse all stocks") rather than a rank-date snapshot from this simulation, so today's candidates always carry the identical qte77 Score as their source universe (ADR-0014).

Series B (reconstructed backfill). Hypothetical, point-in-time backtest — not investment advice. Long the top 25 / short the bottom 25 of all base universes, ranked by a reduced point-in-time score (score_bt, ADR-0013) computed only from what was knowable on each historical rank date. Equal-weighted 1/25 per leg (gross 200 %, net 0), marked daily from closing prices, rebalanced on five cadences (monthly is primary), costed at 10 bp one-way turnover, published gross and net. Returns are total returns (dividends included) in local currency; survivorship and a "today's index membership" size bias are disclosed in each series' caveats under "How it's tested". Extended forward every Saturday.

Each series' caveats are listed under How it's tested.

Design decisions

• ADR-0005 — sentiment and risk sources
• ADR-0006 — the federal-contractors universe
• ADR-0011 — SPY as the equity overlay
• ADR-0012 — the long/short model portfolio
• ADR-0013 — the point-in-time backtest (series A/B)
• ADR-0014 — the latest picks use the live Score
• All decisions