Criterica Intelligence — production models trained on real court records, not synthetic data
For Funders

Judge Behavior as an Underwriting Input

A judge behavioral profile is a record of docket-management and ruling patterns, built from that judge's own rulings — not a prediction of how any single pending case will be decided. The distinction determines how underwriting should use it.

What a behavioral profile actually contains

A useful judge behavioral profile is built entirely from that judge's own public rulings and is descriptive, not predictive of a specific case's outcome: rates at which the judge grants or denies motions to dismiss and summary judgment, by claim type; median time from filing to ruling on dispositive motions; rates of granting or denying class certification where applicable; scheduling patterns (how strictly deadlines are enforced, how often continuances are granted); and, where public record permits it, patterns in damages awards or settlement-approval scrutiny in cases that reached that stage. Every element in the profile should be traceable to specific prior rulings by that judge — it is a compiled record, not an inference from demographic or biographical proxies.

What a properly built profile does not contain is any claim about how a specific pending case in front of that judge will be decided. Judge behavior is one input among several — case facts, claim strength, and party conduct still dominate any individual outcome. The profile answers "how does this judge typically handle motions of this type" and "how long does this docket typically take," not "will this judge rule for the plaintiff here."

How underwriting should use it

The direct underwriting use is procedural, not substantive: adjusting the expected duration and the probability that a case survives a specific dispositive motion, based on that judge's documented rate of granting or denying motions of that type in similar cases. A judge with a documented low rate of granting motions to dismiss in a given claim category changes the expected value of "surviving early dismissal" relative to the circuit average — a real, defensible adjustment grounded in the judge's own record, not a character judgment about the judge.

It is equally useful for what it rules out: docket-management patterns (continuance frequency, average time between filings and rulings) feed the duration distribution directly, independent of the substantive outcome question. A fund managing capital-deployment timing benefits from knowing that a case is in front of a judge whose docket typically runs long, regardless of how the case is likely to be decided on the merits.

Where the line is, and why it matters to hold it

The failure mode to avoid is treating a behavioral profile as a proxy for predicting an individual ruling — using aggregate rates as if they applied deterministically to the specific motion in front of the specific judge on the specific facts of one case. Aggregate rates describe a population of past rulings; they do not resolve the specific one pending. A responsible use of judge behavior data states the aggregate rate, states the sample it is drawn from, and lets the underwriter combine it with case-specific facts rather than substituting for them. Providers that present judge data as a single confidence score attached to your specific case are collapsing a population statistic into an individual prediction it cannot support.

Judicial reassignment changes the profile mid-case

A case underwritten against one judge's documented behavioral profile can be reassigned mid-litigation — through recusal, a senior-status transition, retirement, or ordinary docket rebalancing within the court — and when that happens, the underwriting profile built at intake is no longer the operative one. Funds that treat judge assignment as a fixed input recorded once at intake, rather than a field checked periodically against the live docket, can end up carrying a case priced against a judge who is no longer presiding over it, sometimes for months before anyone notices.

This is a specific, checkable monitoring event, not a hypothetical one: docket reassignment is a matter of public record the moment it happens, and building a periodic check for judge-of-record changes into the same monitoring framework used for other case developments closes a gap that is otherwise easy to miss, because reassignment does not generate the kind of dramatic docket entry that draws attention the way a ruling on a dispositive motion does.

The same monitoring discipline should extend to magistrate judge assignments in matters where dispositive motions are referred for report and recommendation before a district judge rules. A magistrate judge's own documented pattern on the specific motion type matters for the interim procedural stage even when the district judge retains final authority, and a change in magistrate assignment can shift the practical odds of a favorable recommendation well before the district judge is ever asked to weigh in.

What to ask for from an intelligence provider

  • 01A judge profile built entirely from that judge's own documented rulings, with the underlying rulings traceable — not an inferred or proxy-based score.
  • 02Sample size and time window behind every rate reported, so a rate built from five rulings is not presented with the same confidence as one built from five hundred.
  • 03A clear separation between docket-management metrics (duration, continuance rate) and substantive ruling rates, since they feed different parts of the underwriting model.
  • 04An explicit statement of what the profile does not claim — specifically, that it does not predict the outcome of any individual pending motion.
  • 05A defined process for updating the underwriting file when the assigned judge or a referred magistrate judge changes mid-case, rather than treating judge assignment as a fact fixed permanently at intake.

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.

Ask about judge-behavior data in a custom model.

Talk to Custom Models
← All Funder Guides