Metrics interpretation guide
How to read a judge profile. Every figure links to its source.
A commercial matter, listed before a Supreme Court judge.
- Timing: time from hearing to judgment, beside the court's figure. Same-day decisions are included, so open the judgments table to see the spread.
- Outcomes: open the Commercial tab and read granted, partially granted, refused and dismissed, with the sample size and confidence interval underneath.
- Who appears: the barristers and firms the judge sees most. Each name opens an appearance briefing.
- Legislation and authorities: what the judge cites most. Select one to narrow the judgments table to the decisions that cite it.
- Verify: every row links to the full judgment on NSW Caselaw.
For the other side, use the counsel briefing: appearances by court, side and judge. Counts only; no outcome rates for counsel.
The share of outcome-classified decisions in which the application succeeded. 45% on 60 classified decisions means about 27 of those 60 judgments record it as granted.
The denominator matters. Many judgments (procedural rulings, sentencing remarks) state no extractable outcome. They are excluded, never guessed. Every rate shows its sample in words ("33 classified decisions") and a 95% confidence interval, with the excluded count as a footnote.
Court baseline: the same-category average across the judge's court. Above it, the judge grants more often than peers; below it, less. Context, not a prediction.
Settlement rate. The share of published matters whose record notes discontinuance or consent orders. It is a floor: most settled matters never produce a judgment. There is no trial rate or party win rate for Australian matters.
The coverage badge shows how much published record sits behind a profile. More decisions, steadier rates.
100 or more decisions. Highly stable.
50 to 99 decisions. Reliable for most purposes.
30 to 49 decisions. Useful for direction; watch the confidence interval.
10 to 29 decisions. Treat as preliminary.
Fewer than 10 decisions. No rate is published: a rate from a handful of judgments would mislead.
Profiles split metrics by matter category (civil, commercial, criminal, employment and others), classified from each judgment's catchwords. A tab recalculates everything for that category, and defaults to your practice area.
Narrower slices carry smaller samples. Re-check the sample size and the coverage badge after switching tabs.
Observed tendencies: representation mix (legal aid, self-represented, private counsel), judgment types (principal or interlocutory), citation influence and legislation. Citation influence counts later decisions in the corpus that cite the judge's judgments, as a percentile: not adjusted for time on the bench, and not a quality score.
Costs. Each NSW judgment row shows whether costs were ordered and on what basis (ordinary, indemnity, no order), read from the orders. Judge-level costs aggregates are withheld pending legal review.
Rule-based, computed from the published record. No generative AI. Outcome classification accuracy is measured on a reviewed sample and published on the methodology page.
History only. CASE IT never predicts, scores or ranks judges, and every figure leads back to its judgments.