feat(transform): opportunity score v4 — market validation + population-weighted aggregation

Two targeted fixes for inflated country scores (ES 83, SE 77):

1. pseo_country_overview: replace AVG() with population-weighted averages
   for avg_opportunity_score and avg_market_score. Madrid/Barcelona now
   dominate Spain's average instead of hundreds of 30K-town white-space
   towns. Expected ES drop from ~83 to ~55-65.

2. location_profiles: replace dead sports culture component (10 pts,
   tennis data all zeros) with market validation signal.
   Split scored CTE into: market_scored → country_market → scored.
   country_market aggregates AVG(market_score) per country from cities
   with padel courts (market_score > 0), so zero-court locations don't
   dilute the signal. ES (~60/100) → ~6 pts. SE (~35/100) → ~3.5 pts.
   NULL → 0.5 neutral → 5 pts (untested market, not penalised).

Score budget unchanged: 25+20+30+15+10 = 100 pts.
No new models, no new data sources, no cycles.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Dockerfile

@@ -26,6 +26,7 @@ RUN mkdir -p /app/data && chown -R appuser:appuser /app
 COPY --from=build --chown=appuser:appuser /app .
 COPY --from=css-build /app/web/src/padelnomics/static/css/output.css ./web/src/padelnomics/static/css/output.css
 COPY --chown=appuser:appuser infra/supervisor/workflows.toml ./infra/supervisor/workflows.toml
+COPY --chown=appuser:appuser content/ ./content/
 USER appuser
 ENV PYTHONUNBUFFERED=1
 ENV DATABASE_PATH=/app/data/app.db

transform/sqlmesh_padelnomics/models/serving/location_profiles.sql

@@ -16,7 +16,7 @@
 --   10 pts  economic context  — income PPS normalised to 200 ceiling
 --   10 pts  data quality      — completeness discount
 --
--- Padelnomics Opportunity Score (Marktpotenzial-Score v3, 0–100):
+-- Padelnomics Opportunity Score (Marktpotenzial-Score v4, 0–100):
 --   "Where should I build a padel court?"
 --   Computed for ALL locations — zero-court locations score highest on supply gap.
 --   H3 catchment methodology: addressable market and supply gap use a regional
@@ -26,7 +26,9 @@
 --   20 pts  economic power     — income PPS, normalised to 35,000
 --   30 pts  supply gap         — inverted catchment venue density; 0 courts = full marks
 --   15 pts  catchment gap      — distance to nearest padel court
---   10 pts  sports culture     — tennis courts within 25km
+--   10 pts  market validation  — country-level avg market maturity (from market_scored CTE).
+--                               Replaces sports culture proxy (v3: tennis data was all zeros).
+--                               ES (~60/100) → ~6 pts, SE (~35/100) → ~3.5 pts, unknown → 5 pts.
 --
 -- Consumers query directly with WHERE filters:
 --   cities API:       WHERE country_slug = ? AND city_slug IS NOT NULL
@@ -130,8 +132,8 @@ with_pricing AS (
   LEFT JOIN catchment ct
     ON b.geoname_id = ct.geoname_id
 ),
--- Both scores computed from the enriched base
+-- Step 1: market score only — needed first so we can aggregate country averages.
-scored AS (
+market_scored AS (
   SELECT *,
     -- City-level venue density (from dim_cities exact count, not dim_locations spatial 5km)
     CASE WHEN population > 0
@@ -180,8 +182,24 @@ scored AS (
           END
       , 1)
     ELSE 0
-    END AS market_score,
+    END AS market_score
-    -- ── Opportunity Score (Marktpotenzial-Score v3, H3 catchment) ──────────
+  FROM with_pricing
+),
+-- Step 2: country-level avg market maturity — used as market validation signal (10 pts).
+-- Filter to market_score > 0 (cities with padel courts only) so zero-court locations
+-- don't dilute the country signal. ES proven demand → ~60, SE struggling → ~35.
+country_market AS (
+  SELECT
+    country_code,
+    ROUND(AVG(market_score), 1) AS country_avg_market_score
+  FROM market_scored
+  WHERE market_score > 0
+  GROUP BY country_code
+),
+-- Step 3: add opportunity_score using country market validation signal.
+scored AS (
+  SELECT ms.*,
+    -- ── Opportunity Score (Marktpotenzial-Score v4, H3 catchment) ──────────
     ROUND(
       -- Addressable market (25 pts): log-scaled catchment population, ceiling 500K
       25.0 * LEAST(1.0, LN(GREATEST(catchment_population, 1)) / LN(500000))
@@ -195,10 +213,14 @@ scored AS (
           END, 0.0) / 8.0)
       -- Catchment gap (15 pts): distance to nearest court
       + 15.0 * COALESCE(LEAST(1.0, nearest_padel_court_km / 30.0), 0.5)
-      -- Sports culture (10 pts): tennis courts within 25km
+      -- Market validation (10 pts): country-level avg market maturity.
-      + 10.0 * LEAST(1.0, tennis_courts_within_25km / 10.0)
+      -- Replaces sports culture (v3 tennis data was all zeros = dead code).
+      -- ES (~60/100): proven demand → ~6 pts. SE (~35/100): struggling → ~3.5 pts.
+      -- NULL (no courts in country yet): 0.5 neutral → 5 pts (untested, not penalised).
+      + 10.0 * COALESCE(cm.country_avg_market_score / 100.0, 0.5)
     , 1) AS opportunity_score
-  FROM with_pricing
+  FROM market_scored ms
+  LEFT JOIN country_market cm ON ms.country_code = cm.country_code
 )
 SELECT
   s.geoname_id,

transform/sqlmesh_padelnomics/models/serving/pseo_country_overview.sql

@@ -18,13 +18,14 @@ SELECT
   country_slug,
   COUNT(*)                                                        AS city_count,
   SUM(padel_venue_count)                                         AS total_venues,
-  ROUND(AVG(market_score), 1)                                    AS avg_market_score,
+  -- Population-weighted: large cities (Madrid, Barcelona) dominate, not hundreds of small towns
+  ROUND(SUM(market_score * population) / NULLIF(SUM(population), 0), 1) AS avg_market_score,
   MAX(market_score)                                              AS top_city_market_score,
   -- Top 5 cities by venue count (prominence), then score for internal linking
   LIST(city_slug ORDER BY padel_venue_count DESC, market_score DESC NULLS LAST)[1:5]     AS top_city_slugs,
   LIST(city_name ORDER BY padel_venue_count DESC, market_score DESC NULLS LAST)[1:5]     AS top_city_names,
-  -- Opportunity score aggregates (NULL-safe: cities without geoname_id match excluded from AVG)
+  -- Opportunity score aggregates (population-weighted: saturated megacities dominate, not hundreds of small towns)
-  ROUND(AVG(opportunity_score), 1)                               AS avg_opportunity_score,
+  ROUND(SUM(opportunity_score * population) / NULLIF(SUM(population), 0), 1) AS avg_opportunity_score,
   MAX(opportunity_score)                                         AS top_opportunity_score,
   -- Top 5 opportunity cities by population (prominence), then opportunity score
   LIST(city_slug ORDER BY population DESC, opportunity_score DESC NULLS LAST)[1:5] AS top_opportunity_slugs,