Methodology & data sources
Built on public court records. What we use, how we count, and what we leave out.
Data sources
Australia
NSW Caselaw
Visit source17,000+ published decisions from five NSW courts (Supreme Court, Court of Appeal, Court of Criminal Appeal, District Court, Land and Environment Court), January 2021 to August 2026.
Metrics derived: Judge attribution, application outcomes, hearing and decision dates, catchwords, matter categories, legislation and authorities cited, counsel appearances, costs orders
Open Australian Legal Corpus
Australian federal and state court decisions. The basis for decision statistics on about 2,000 Australian judicial officers.
Metrics derived: Decision counts, matter categories, citation counts, appellate records where linked
Coverage statistics
- Australian judicial officers
- 1,970
- Judges with metrics (10+ decisions)
- 1,082
- NSW decisions
- 17,322
- AU courts with judges
- 29
- Court divisions
- 30
A US federal dataset (4,500+ district judges, 11M+ FJC case records) is maintained and returns with regional expansion.
How metrics are calculated
Application outcome rates
The share of classified published decisions in which the application was granted, partly granted or refused. A decision with no clear disposition is left unclassified and excluded from every rate, and the profile shows how many were excluded. Coverage levels by sample: very high (100+), high (50–99), medium (30–49), low (10–29). Rates on fewer than 10 matters are not shown. Every rate carries its count and 95% confidence interval.
Settlement & trial rates
Australian settlement rates come from the published record (settlement, discontinuance and consent orders in catchwords and orders). There is no Australian trial rate or party win rate: most settlements produce no judgment, and orders rarely name a winner.
Costs orders
Read from the orders of each NSW judgment: whether costs were awarded, and on what basis (ordinary, indemnity, no order). Measured on a held-out labelled set: 98.3% precision and 100% recall for the award, basis correct in every case. Shown per judgment; judge-level costs aggregates are withheld pending legal review.
Counsel and firm appearances
Read from the representation block of each NSW judgment, with side and seniority (SC, KC, QC) where stated. Matched on the published appearance name only. Self-represented parties are never recorded as counsel. Reported as counts of published judgments by court, side and judge. No counsel-level outcome rates are calculated or published.
Case duration & decision time
Australian decision time is hearing to judgment, where a hearing date is published. Each figure sits beside the court's own.
Citation influence
Citations of a judge's judgments by later decisions in the corpus, as a count and a percentile. Not adjusted for time on the bench or volume, so long-serving judges accumulate more. It describes citation frequency only. It is not a quality score or a ranking.
Procedural patterns (AI insights)
Written by a rule-based engine from the figures above. No external AI service is used. The wording describes historical patterns only: no advice, no predictions.
Limitations & disclaimers
Coverage varies by judge. Some judges have too little published history for a rate. The coverage badge on each profile says how much there is.
Historical patterns only. Past patterns do not predict future outcomes. Every decision turns on its own facts.
Coverage is jurisdiction-specific. Judgment records cover NSW courts, refreshed monthly. Other Australian courts carry decision statistics from the Open Australian Legal Corpus.
Measured classifier accuracy (NSW outcomes). Outcomes are read from the published orders. A decision with no clear disposition is left unclassified and excluded from every rate. In a reviewed stratified random sample of 210 decisions (August 2026), each label was compared with the operative orders: 94.5% carried the exact outcome class (95% CI 90–97%) and 99.5% the correct direction (95% CI 97–100%). On the rest, the classifier abstained. The review was AI-assisted and has not yet been independently verified by a legal practitioner; that verification is scheduled. The sample is frozen in our test suite and re-measured whenever the classifier changes.
Our commitments. History only: no predictions, scores or rankings. Every figure links to its public source. Rates on fewer than 10 matters are not shown; every rate carries its sample size and 95% confidence interval. Public professional record only: no personal or lifestyle data. Judge-level family-law metrics are withheld pending legal review. Corrections: hello@caseit.legal.
Not legal advice. CASE IT is a research tool. It gives no legal advice, strategy or predictions. Litigation decisions belong with a qualified legal practitioner.
Full metric definitions: metrics interpretation guide.