San Francisco judges by release type
Each table ranks San Francisco judges by the share of people they released in that category who came back with a new case. The three release types involve different decisions and different defendants, so each gets its own table.
Data runs through May 1, 2026. It comes from a public dataset published on May 4, 2026 that has not been updated since, so releases and new cases after that date are not counted.
- Three tables, not one. An earlier version of this page mixed all three release types into a single ranking, which made Mental Health Court judges look "lenient" relative to arraignment judges. That was a false comparison; they're doing different jobs. Below, each table compares like to like.
- 365-day window is the primary number. A judge whose releases are all from 2023 has had 2-3 years for those defendants to potentially reoffend. A judge whose releases are mostly from 2025 has 6-12 months. To level the playing field, the bold number for each judge below is the reoffense rate within 365 days of release, counted only for releases old enough to have had that full window. The all-time number is shown in lighter type beside it as a sanity check.
Own Recognizance releases
Defendant released without bail at arraignment. Most common SF release decision.
Mental Health Diversions granted
PC 1001.36 diversion into the SF Behavioral Health Court program. Defendants are screened before admission; reoffense here usually means the diversion was terminated or the defendant picked up new charges while in treatment.
PC § 1385 dismissal cases (Interest of Justice)
Judge dismisses one or more charges. The DA may oppose; the judge does it anyway. One case may have multiple charges dismissed; we count it once.
Methodology
Release event = an Agreement and Order Releasing Defendant on Own Recognizance, a mental-health diversion granted under Penal Code § 1001.36, or a charge dismissal under § 1385 (Interest of Justice). Each event is counted at the case level (one per case), not the charge level. A case with multiple § 1385 dismissals counts once.
Judge attribution = for each release event, the SF Superior Court judge whose calendar entry on this case was nearest in time to the event date (within 90 days). About 95% of release events attribute under this rule; the rest appear in each table as an "Other (unattributable)" row.
Reoffense = the same defendant has another SF criminal case filed against them after the release date. Conservative version: re-charging, not re-conviction.
365-day window = the primary rate. To level the playing field between a judge whose releases are mostly from 2023 (3 years of follow-up) and a judge whose releases are mostly from 2025 (less than a year), we only count releases that happened on or before 2025-05-01 (the dataset's effective cutoff minus 365 days) and only count reoffenses that happened within 365 days of release. The all-time rate is still shown in lighter type as a comparison point. When it's much higher than the 365d rate, that judge has a long tail of late reoffenses; when it's close, their rate is stable across time.
Sample thresholds: a judge only appears in a table if they have ≥30 OR releases, ≥30 MH diversions, or ≥15 PC § 1385 dismissal cases respectively (counting all-time, not just eligible). Each table notes how many cases sit below that threshold.
Source: jamiequint/sf_criminal_court (CC-BY-NC 4.0).