Reverse liquidity stress as a run-off multiplier with a refusal rule, scored against a labelled run

Validating a liquidity stress model needs the inverse of the forward scenario: the run-off speed at which a stated survival period is reached. We define the reverse liquidity stress as the smallest multiplier on the daily run-off rates, capped so that no rate exceeds one, at which the survival period is at most a target. It is the root of a continuous window function, rounded up with a neighbour check; a target that no multiplier brackets is refused with its reason and no number. Seven propositions with proofs give its monotonicity, refusal rule, exact round trip and sensitivity. On 20,000 seeded instances the engine agrees with an independent oracle on every outcome and returns a threshold on every solved one; a case-grid scan answers 93.5 (93.0 to 93.9) % of the refused instances and overshoots by a median of 21.7 %. On case C16 the multiplier for 30 days is 2.118175 (LiST Adverse) and 1.383141 (LiST Extreme), and the 2023 fast-run base is refused. Calibrated on rows before 2023, that base models 0.1708 of deposits in the two days of the March 2023 run against 0.85 observed (ratio 0.201), and the multiplier that reproduces the run is 12.22 times the calibrated rate. The Adverse multiplier moves by 49.8 % over the joint band of case conventions: under the slow monetisation schedule it is refused at the central haircut and falls to 1.063646 at the low haircut. Nine of twelve predictions held; three failed. The bank's deposit behaviour is synthetic. Repository and reproducible pipeline (MIT): https://github.com/fsantibanezleal/CAOS_Contraste .