About this terminal

This is a working market-based credit-rating system: a KMV/Merton structural model estimated by EM over five years of equity data, converted to Time-Consistent (TiC) credit measures and an S&P-equivalent letter, run across a 150-name universe chosen to include the hard cases — banks, ADRs, negative book equity, recent IPOs, genuinely distressed names. Everything on this site was computed offline from committed data and is reproducible from the repository with two commands.

Finding one — the letter is the weak layer

The drift-free RiskScore inherits exactly the volatility's uncertainty (×2.00, the algebraic square) while the PD-based letter conversion amplifies it roughly ×4,073 on the measured universe. The same letter also swings up to seven notches on an unargued debt-weight convention, and for deposit-funded banks no convention lands on the truth at all. The rank ordering underneath is robust to all of it — which is why this site shows RiskScore first and never renders a letter without its interval.

Finding two — validated, honestly

Against sourced agency ratings the ordering discriminates well: Spearman ρ = 0.79 overall and 0.73 restricted to the names where the scale genuinely resolved a letter, holding within every sector. Calibration is the failure mode: the letters run a median five notches optimistic, and plain distance-to-default ties the TiC RiskScore on discrimination — a result we report rather than bury. The full study, including where the model fails, is in the repository.