Crime Watch DC

1. Primary incident files

We downloaded the official yearly incident extracts:

• Crime Incidents in 2026 — yearly extract used to build the embedded map file (13,563 geocoded pins through Aug 27, 2026). Live YTD on the snapshot bar can run ahead of that extract.
• Crime Incidents in 2025 — 24,149 records in the downloaded CSV; live FeatureServer layer 7 has 24,151. MPD year-end is 24,177

Publisher: Metropolitan Police Department via Open Data DC / ArcGIS Hub. License: CC BY 4.0.

2. Official year-end and YTD tables

Completed-year totals (2024 and 2025, including homicides) are taken from MPD’s published year-end comparison tables on District Crime Data at a Glance. Those tables use DC Code offense definitions and are preliminary. They are not FBI UCR or NIBRS official statistics. Cases can be reclassified later.

2026 YTD and the matching 2025 window are not scraped from that page. They come from grouped Open Data queries on the official yearly FeatureServer layers (2026 layer 41, 2025 layer 7), using the same DC Code offense names. DC Crime Cards is the closest official dashboard to check against, because it also uses report date. Use Verify vs official on the Stats tab to re-query that file, show a category-by-category match, and write the live totals from it.

3. How we categorize offenses

• Homicide → Homicide
• Assault w/ Dangerous Weapon → Assault
• Robbery → Robbery
• Burglary → Burglary
• Theft/Other + Theft F/Auto → Theft
• Motor Vehicle Theft → Auto
• Sex Abuse → Sex Abuse (violent)
• Arson → Arson (property)

Violent and property totals follow MPD’s published DC Code index: violent = homicide + sex abuse + assault w/ dangerous weapon + robbery; property = burglary + theft (other + F/auto) + motor vehicle theft + arson. The map filter still groups sex abuse and arson on one checkbox so the pin list stays compact.

Year-to-date totals and the 2025 matching window use REPORT_DAT (the same clock as Crime Cards and MPD Daily Crime). Hour-of-day, weekday, and shift charts still use START_DATE, the offense clock. The monthly comparison charts on the Stats tab still use the embedded extract’s month buckets (offense start month in the downloaded file) and only run through the extract cutoff, so they can lag the live YTD ribbon. Homicide official tables use date declared and can differ by a case or two.

4. What the map shows

* All Months * with every crime type and ward checked is citywide YTD from the Open Data summary query (report date). If you uncheck a ward or category, or pick a month or time of day, the snapshot bar and sidebar counts switch to the visible map points and are labeled Filtered. They will not match MPD’s citywide table while a filter is on.

The Map tab plots every 2026 geocoded incident in the embedded extract — 13,563 points with valid coordinates — then replaces dates inside the live 14-day window with the FeatureServer pull. A handful of rows lack usable lat/lng and are omitted. Addresses are already anonymized to the block by MPD. Ward, category, and time filters change which pins draw. Clusters stay on so the browser can handle the full set.

5. Snapshot time and the live overlay

The green line on the map is only Snapshot [date] | [time]. That stamp is the latest REPORT_DAT in the official 2026 Open Data file (Eastern time), not the minute you opened this page. MPD’s public file usually runs through the prior evening, not the current hour.

Map pins keep the embedded extract for older days. The last 14 report dates are pulled from the MPD FeatureServer and replace extract pins in that window so recent blocks stay current without refetching all 13k points. With every filter off, year-to-date totals on the snapshot bar come from a grouped count query (offense totals + max report time), not from re-downloading every point. If a ward, category, month, or time filter is on, that bar switches to a count of the visible pins and is labeled Filtered.

Refresh now in the footer can run only once an hour. The page also auto-checks once an hour while the tab is visible.

6. Fair 2026 comparison

2026 is incomplete. The Stats tab compares 2026 YTD with the same calendar window in 2025 (Jan 1 through the current 2026 as-of report date), not with the full 2025 year. Both sides of that comparison use REPORT_DAT. Use Update snapshot on Stats to refresh both sides. Full-year 2024 vs 2025 remains the cleanest completed-year comparison.

7. Experimental dashboard

CrimeWatchDC is an experimental dashboard currently in development. Figures can lag, change after reclassification, or differ from an agency’s own published table. For 100% official accuracy, check the Metropolitan Police Department and District of Columbia government sites, including District Crime Data at a Glance, DC Crime Cards, and Open Data DC.

8. What updates live, what is reviewed monthly

• Map pins. Embedded 2026 extract for older days. Last 14 report dates overlay from FeatureServer 41, hourly while the tab is visible (manual refresh once an hour).
• Snapshot bar and Stats 2026 YTD. Citywide totals come from a grouped Open Data count query (offense totals + latest REPORT_DAT), not a 13k-point download. A filter on the map rebuilds the snapshot bar from visible pins instead. Stats → Update snapshot also rebuilds the matching 2025 REPORT_DAT window from FeatureServer 7.
• Facts cards and In Focus count tiles. Rewritten from that same live rollup after each successful pull, so homicide / assault / auto-theft sentences are not frozen to the extract day.
• Last 7 Days tab. Rolling 7-day slice of the live 14-day REPORT_DAT window. The heading dates update with the pull.
• In Focus tactics. Place-and-hour notes. Reviewed monthly. A quieter week does not rewrite the corridor play.
• Theft F/Auto vs Theft/Other split on In Focus. Extract split only — the live feed still folds both into Theft.
• 2024 / 2025 year-end tables. Published MPD completed-year figures. They do not tick hourly.

Embedded extract through Aug 27, 2026. Live windows print in the footer after the first successful hourly check.