# North Hoboken Family Map — Answer

Open **`index.html`** (double-click; needs internet only for Leaflet CDN + map tiles). No build, no keys.

## What's on the map (27 real places)

- **Playgrounds (5):** Church Square Park, Columbus Park, Southwest Resiliency Park, Elysian Park, Pier C Park.
- **Family food (5):** Benny Tudino's, Fiore's House of Quality, Carlo's Bake Shop, O'Bagel, Lisa's Italian Deli.
- **Activities (5):** Hoboken Historical Museum, Fire Dept Museum, Liberty Science Center, Pier 13, Pier A Park.
- **Day trips (4):** Central Park, AMNH, Liberty State Park, Bronx Zoo.
- **Date night (5, added per prompt 2):** Sorellina, Augustino's, Amanda's (BYOB), Otto Strada, Halifax.
- **Tapas (3, added per prompt 3):** Bin 14 (Hoboken wine bar with explicit small-plates tapas menu), Tapas de España (Heights, Jersey City), La Isla (Cuban, shareable-plates adjacent). Note: Hoboken proper has almost no traditional Spanish tapas house; Bin 14 is the honest in-town pick. "El Ocho" was researched and does not appear to exist — excluded rather than invented.

All places verified against OpenStreetMap/official sites; addresses in `places.csv`.

## Hidden-gem score (prompt 4)

- Method pre-registered in `METHODOLOGY.md` **before** scoring.
- Formula: `gem = 100 × quality × obscurity`, quality from shrinkage-adjusted Google rating, obscurity from log review count. Inspectable: `places.csv` (data) → `score.py` (script) → `scores.json`/`scores.csv` (outputs) → `VALIDATION.md` (test).
- **Validation (locked priors): Spearman ρ = 0.956 (criterion > 0.7); pairwise 11/12 (criterion ≥ 10); top-3 and bottom-3 exact.** The single miss: Elysian vs Columbus (36.9 vs 37.0) — near-tie, directionally wrong by 0.1 points.
- Top gems: Fire Dept Museum (48.3), Southwest Park (47.1), Lisa's Deli / Historical Museum (40.5). Landmarks correctly bottom: AMNH (0.4), Central Park (0.0).
- Honesty note: 6 Google rating/counts are aggregator-mirrored (flagged `verified-mirror`); the rest are labeled `estimate`/`estimate-order-of-magnitude` in `places.csv`. Replace with live Google values and re-run `python3 score.py` — methodology needs no change.