Single-Trajectory Local Filtering (STLF): Noise Rejection and Signed-Triad (−1,0,+1) Ledger Control for Missingness and Instability

This article introduces Richardson Single-Trajectory Local Filtering (STLF) as a ledger-first upgrade to standard single-copy local filtering (ScLF) for noisy entanglement tests and other platforms where drift, missingness, and low-yield “wins” can silently distort conclusions. The central change is bookkeeping: the experiment is treated as one sequenced trajectory, every attempt is assigned a global index, and each of three partial channels writes a symbol in at that index. Here are signed hits and is a literal no-hit/no-decision outcome that must remain visible; the rule is strict: zeros stay zeros (no smoothing, skipping, or post-hoc repair). The second change is architectural: STLF replaces the usual two-partial approach with a triad of partials so that outliers, drift, and “fake dominance” (direction rises only because yield collapses into zeros) can be detected and charged rather than hidden. Direction and yield are tracked separately per partial and combined only by an explicit, replayable scoring rule, so a channel cannot claim improvement by going silent. The protocol also specifies falsifiable requirements: the ledger must be replayable, knob adjustments must be declared functions of the written ledger, and the third partial must remain an independent witness rather than a deterministic echo of the other two. STLF is framed to map cleanly into corridor-style containment and token workflows, and it is explicitly designed to work alongside related corridor tools (ternary knob certification and signed-cycle/Strong-to-Weak tagging) by enforcing the same “zeros as first-class outcomes” accounting discipline.