Show HN: Play with an interactive heatmap of SF crime (and other cities)

Metrics, Risk, and Population Density

• Many argue raw crime counts mostly reflect where people are, not per-person risk; per-capita normalization is urged to avoid “just mapping population density.”
• Others say the “right” metric depends on use case:
  • Traveling through an area: interest in crimes per area or per hour.
  • Living somewhere: crimes per resident, esp. home-targeting crimes.
• Several note that daytime / tourist / worker population can diverge from resident population; ideal denominator might be “person-hours in area,” which is largely unavailable.
• Some dispute simplistic “more density = more risk,” pointing to high-density but low-crime environments and social-network-driven victimization patterns.

Tourism, Car Break‑Ins, and Practical Safety

• Tourists debate whether such maps are useful for short trips; some prefer common sense, others want to avoid hotspots, especially for theft of cars, phones, and documents.
• Car break-ins cluster at tourist areas like Fisherman’s Wharf and scenic spots; advice includes not leaving valuables visible, avoiding tourist traps, sometimes even leaving cars unlocked and empty.
• Tool is seen as especially valuable for housing searches and understanding neighborhood-level issues (e.g., Tenderloin, Mission, prostitution corridors).

Data Quality, Biases, and Interpretation

• Concerns about:
  • Crimes geocoded to police HQ or reporting offices rather than actual locations.
  • Under‑ or non‑reporting, changing reporting standards, and political incentives to manipulate statistics.
  • Crime maps effectively visualizing enforcement/reporting patterns, not true incidence (“WWII plane” analogy).
• Examples cited of crime rates distorted by tourists, transient populations, or outlier events.

UI, Features, and Visualization Choices

• Praise for fast, slick UX, clear labels, and OSM base map.
• Requested features: per-capita toggle, longer historical ranges, trend/delta maps, city comparisons, customizable crime groupings, color-coding combinations, and user‑saved configurations.
• Suggestions for auto-hiding side panels, handling map bounds, and improving low-count visualization (dots, clustering, DBSCAN instead of continuous KDE).
• Ideas for alerts (“danger zone” notifications), positive-data maps (views, kindness), and expansion to more US and European cities.