\subsection{An explicit certificate for the endpoint-score residual} \label{sec:finite-residual} The quadratic bound $B_n$ can also be supplied with a finite upper companion. The certificate concerns the exact stationary endpoint experiment. It identifies the residual with a conditional binomial score, then bounds that score by a polynomial with an exactly computable second moment. The conditional-score principle is classical \cite{Louis1982}, as is the Krawtchouk--Ehrenfest framework \cite{DiaconisGriffiths2012}; the required normalization and identities are derived here. With the endpoint pair $Y$, the statistics $A,D$, the endpoint log-clock score $S$, the centered quadratics $W_1,W_2$ and the projection $\Pi_2$ of Section~\ref{sec:finite-certificate}, define the endpoint-score residual \begin{equation}\label{eq:residual-E} E_n(t)=\E[(S-\Pi_2S)^2] =\lambda\{R_n(\lambda)-B_n(t)\},\qquad\lambda=nt. \end{equation} For the following formulas, put \begin{align} s&=1-p,&r&=1-2p=e^{-2t},\notag\\ w&=ps,&\nu&=nw,\notag\\ h&=p^2+s^2=1-2w,& \gamma&=\frac{tr}{w}=\frac{2t}{\sinh(2t)},\notag\\ d&=2\nu+r^2,&e&=2\nu+3-14w, \label{eq:residual-parameters} \end{align} where $p=(1-e^{-2t})/2$. Define the polynomial \begin{equation}\label{eq:residual-V} \mathcal V_n(p)=2n(n-1)w^2 \left[h^2+4hr^2+(n-2)wh+(11n-14)wr^2\right] \end{equation} and the explicit certificate \begin{equation}\label{eq:residual-T} T_n(t)=\frac{\gamma^2r^2\mathcal V_n(p)}{d^2e^2}. \end{equation} \begin{theorem}[Finite residual and relative information bounds] \label{thm:finite-residual} Let $R_n$, $B_n$, $E_n$ and $T_n$ be as in \eqref{eq:R}, \eqref{eq:finite-endpoint-bounds}, \eqref{eq:residual-E} and \eqref{eq:residual-T}, with $\lambda=nt$. For every integer $n\ge2$ and $0