What the site includes
The database is built from the public jamiequint/sf_criminal_court dataset. This site narrows that source to charge dispositions with no jail or prison days listed and an outcome categorized as diversion, dismissal, acquittal, deferred judgment, or discharge.
Records are excluded when the disposition indicates another active path, such as a guilty plea to other charges, federal indictment, release to another agency, violation proceedings, or a deceased defendant.
What the site does not prove
An eligible record does not prove that nothing happened in the full case. It means this charge record matched the site’s narrow outcome and sentence filters. A case can contain multiple charges, later activity, or related proceedings that are outside this app’s eligible pool.
Judge attribution
Judge names are attached by exact case/date matches to the dataset’s published department assignments. Placeholder assignments such as unknown, visiting, vacant, or TBD judges are removed.
A department assignment is not guaranteed to identify the actual judicial officer who made every decision at a hearing. Treat judge-level summaries as attribution based on the source dataset’s assignment fields, not as a complete causal record.
Defendant names
Defendant ratings use normalized names from the dataset. They are not durable person identifiers. Same-name collisions and spelling variants can happen, so defendant leaderboards should be read with extra caution.
How Elo ratings work
The main page shows two eligible charge outcomes. When a visitor picks the charge that sounds worse, the judge and defendant attached to that selected charge receive an Elo win against the other side. Higher ratings mean site voters have more often selected worse-sounding eligible charges attached to that entity.
Elo ratings are a crowd signal from this website. They are not official court statistics, legal findings, or a complete measure of a judge, defendant, case, or charge.
How to use the dashboard
The explore page is meant to show the shape of the eligible pool before anyone interprets a ranking. Compare ratings with vote counts and eligible record counts. Small samples can move quickly, and common outcomes or statutes can dominate what voters see.