bonds — an honest lens on the distressed universe

A screening & data-discipline tool for the deep-distress corporate-bond cohort (entry price ~25–40 of par). It reports what the cohort did and points you at the factors worth investigating. It is not investment advice and makes no per-name prediction.

NOT A CALL. Per-name recovery is not forecastable on free data (out-of-sample AUC ≈ 0.44, below chance in deep distress). The tool therefore abstains on per-name calls and instead shows the cohort base rate — always with its survivorship ledger — and the gated factors to investigate.

What it gives you

• Membership — is a bond in the deep-distress cohort as of a date?
• Cohort base rate — recover / default / still-distressed at 18m, computed point-in-time (only from episodes that had resolved before that date), shipped with an explicit coverage/survivorship ledger. It is an upper bound on recovery.
• Surviving flags — only factors that passed an out-of-sample, deflation-corrected, regime-neutralized eval are shown (currently delisting and fallen-angel). Going-concern (null eval) and capital-structure-floor (not built) are visibly excluded.

Two modes

Current = latest available panel (no free real-time per-bond feed exists). History = pick a past date and see exactly what the tool would have said then, under point-in-time gates (no look-ahead, no back-filled prices, no outcomes).

API

• GET /api/base-rate?date=YYYY-MM-DD — cohort base rate as-of a date (omit date for current)
• GET /api/cohort?date=YYYY-MM-DD — membership + gated flags for the as-of-date cohort
• GET /api/card?cusip=XXXXXXXXX&date=YYYY-MM-DD — one bond's honest card
• GET /api/meta · GET /healthz

No hypothetical or backtested performance is shown; realized-vs-baseline evaluation is internal QA only. This is general information, not personalized investment advice (MiFID II / CH FinSA). Data: bond prices via OpenBondAssetPricing (cite; academic); events via SEC EDGAR (public domain); ratings via NRSRO 17g-7.