We don’t predict duration once.
We continuously manage duration as evidence changes.
Duration is the mechanism that converts legal outcomes into financial performance. A position can be right on the merits and still underperform because capital stayed deployed longer than it was priced to stay. Criterica treats time to resolution as a governed, continuously re-estimated variable — a distribution that updates on evidence, carries its own tail, and preserves every prior estimate for audit.
“Expected resolution in 36 months” collapses seven different clocks.
Each clock has its own drivers, its own endpoint, its own data, and its own interventions. A system that models them separately can act on them separately.
Time to the next consequential filing, ruling, or hearing. The shortest clock, and the one that re-times all the others.
Time to a negotiated resolution. Hazard is not constant; it concentrates around procedural pressure points.
Time to a decision on the merits. Runs on the court’s calendar, not the parties’.
A discrete additional layer with its own duration distribution, conditional on the route taken to it.
Judgment to enforceability. For cross-border and sovereign matters, often the dominant clock.
Enforceability to money received. Legal resolution and cash resolution are different dates.
Commitment to redeployment. The clock a portfolio actually compounds on.
The forecast is alive, and its history is preserved.
A matter enters with a baseline: the duration distribution of its true cohort, stated as a band with a monitored tail, never a date. Every consequential event re-conditions that estimate — a survived motion, a certification decision, a scheduling change. The band moves, or holds, and both the movement and the original are kept.
Preserved forecast history is the discipline that separates measurement from storytelling: what the system said at each point is on the record before the outcome is known, so accuracy is auditable and misses are as visible as hits.
A metric family for legal-asset duration.
Credit has ratings. Market risk has VaR. Legal assets have had adjectives. These are the measures Criterica computes and monitors so that duration risk can be stated, compared, and priced.
The 90th-percentile time to resolution. The tail a reserve must carry, stated as a first-class number rather than discovered later.
TaR90 minus the median. The distance between the number a plan is built on and the number one matter in ten actually delivers.
Movement of the duration distribution since underwriting or the last review. The honest record of what changed, preserved against the original.
The probability that a defined share of a book resolves by a given period. Realization planning as a distribution, not a hope.
The share of a book’s time-exposure riding a common driver — one court’s calendar, one doctrine, one procedural regime.
The portion of expected delay that feasible action could remove without damaging position. Defined so that it can be measured, not asserted.
Nobody controls a court. Parts of the clock respond anyway.
The objective is not to control all duration. It is to identify the controllable component earlier, quantify the cost of delay, and measure whether intervention changed the path.
Precision about limits is part of the method.
Duration models are validated on chronological splits — trained on the past, tested on a later period they never saw. Random splits hide regime change, so we do not use them.
Open matters have not finished. Any duration method that quietly drops them inherits a systematic bias toward optimism. Ours accounts for it and says so.
Public dispositions do not reliably encode which party prevailed or what a case was worth. We do not build prevailing-party or case-value models from them, under any name.
A duration estimate is a distribution with a stated construction, a visible tail, and a confidence grade. A single date is false precision, and we do not publish one.
Statistics shown reflect historical or illustrative model outputs derived from real case data. They are not predictions or guarantees of any individual outcome. Litigation results depend on facts, jurisdiction, judge, and counsel, and vary case by case. Model accuracy is subject to selection effects and changing legal dynamics.
The duration research series.
Working papers on duration and its arithmetic consequences for legal-asset returns, published openly as they complete. The series begins with the mechanics every allocator can verify by hand: what time does to an IRR at a fixed multiple.