Describe your matter. Hence weighs it against thousands of verified High Court orders and shows exactly what moves the outcome.
Every estimate is a sum of visible terms. No black box: you can audit each one.
Each model is trained on older orders and scored on the most recent ones. A matter type goes live only when it ranks outcomes well (AUC ≥ 0.65) and beats the historical base rate. Everything else stays in validation.
AUC is the chance the model ranks a successful matter above an unsuccessful one. 0.50 is a coin toss.
No. Benches are assigned by the Chief Justice's roster. Hence shows how the listed bench, or the roster as a whole, has ruled on comparable facts, so you can prepare for whoever hears the matter.
No. Counsel are ranked only on outcomes in comparable matters, measured against what the facts predicted. Hence never models who knows whom, and advocates related to a bench are excluded under the BCI Rules.
Published High Court judgments. Each is read, structured and verified against its own text before it reaches the statistics.
Accuracy is published per matter type above, measured on orders decided after the training period. Matter types that don't clear the bar are not offered.
No. Hence is decision support for advocates. It estimates likelihoods from past orders; it does not predict any individual judge's decision.
Free to use. No sign-up.
data/orders.json to run on real records.