deemanone/padelnomics
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(score): Opportunity Score v7 — calibration fix for saturated markets

Browse Source 
Two fixes:
1. dim_locations now sources venues from dim_venues (deduplicated OSM + Playtomic)
   instead of stg_padel_courts (OSM only). Playtomic-only venues are no longer
   invisible to spatial lookups.
2. Country-level supply saturation dampener on supply deficit component.
   Saturated countries (Spain 7.4/100k) get dampened supply deficit (x0.30 → 12 pts max).
   Emerging markets (Germany 0.24/100k) nearly unaffected (x0.98 → ~39 pts).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Deeman
2026-03-09 18:03:14 +01:00
parent cce3c466ba
commit e39dd4ec0b
2 changed files with 34 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
transform/sqlmesh_padelnomics/models/foundation/dim_locations.sql
View File

@@ -9,7 +9,7 @@
-- foundation.dim_countries → country_name_en, country_slug, median_income_pps
-- stg_nuts2_boundaries + stg_regional_income → EU NUTS-2/NUTS-1 income (spatial join)
-- stg_income_usa → US state-level income (PPS-normalised)
-- stg_padel_courts → padel venue count + nearest court distance (km)
-- foundation.dim_venues → padel venue count + nearest court distance (km)
-- stg_tennis_courts → tennis court count within 25km radius
--
-- Income resolution cascade:
@@ -137,10 +137,12 @@ us_income AS (
PARTITION BY m.admin1_code ORDER BY s.ref_year DESC
) = 1
),
-- Padel court lat/lon for distance and density calculations
-- Padel venue lat/lon for distance and density calculations.
-- Uses dim_venues (deduplicated OSM + Playtomic) instead of stg_padel_courts (OSM only)
-- so Playtomic-only venues are visible to spatial lookups.
padel_courts AS (
SELECT lat, lon, country_code
FROM staging.stg_padel_courts
FROM foundation.dim_venues
WHERE lat IS NOT NULL AND lon IS NOT NULL
),
-- Nearest padel court distance per location (bbox pre-filter → exact sphere distance)

37
transform/sqlmesh_padelnomics/models/serving/location_profiles.sql
View File

@@ -19,19 +19,20 @@
-- 10 pts economic context — income PPS normalised to 25,000 ceiling
-- 10 pts data quality — completeness discount
--
-- Padelnomics Opportunity Score (Marktpotenzial-Score v6, 0–100):
-- Padelnomics Opportunity Score (Marktpotenzial-Score v7, 0–100):
-- "Where should I build a padel court?"
-- Computed for ALL locations — zero-court locations score highest on supply deficit.
-- H3 catchment methodology: addressable market and supply deficit use a regional
-- H3 catchment (res-5 cell + 6 neighbours, ~24km radius).
--
-- v6 changes: lower density ceiling 8→5/100k (saturated markets hit zero-gap sooner),
-- increase supply deficit weight 35→40 pts, reduce addressable market 25→20 pts,
-- invert market validation (high country maturity = LESS opportunity).
-- v7 changes: country-level supply saturation dampener on supply deficit.
-- Saturated countries (Spain 7.4/100k) get dampened supply deficit (×0.30 → 12 pts max).
-- Emerging markets (Germany 0.24/100k) are nearly unaffected (×0.98 → ~39 pts).
-- Floor at 0.3 so supply deficit never fully vanishes.
--
-- 20 pts addressable market — log-scaled catchment population, ceiling 500K
-- 15 pts economic power — income PPS, normalised to 35,000
-- 40 pts supply deficit — max(density gap, distance gap); eliminates double-count
-- 40 pts supply deficit — max(density gap, distance gap) × country dampener
-- 10 pts sports culture — tennis court density as racquet-sport adoption proxy
-- 5 pts construction affordability — income relative to construction costs (PLI)
-- 10 pts market headroom — inverse country-level avg market maturity
@@ -209,17 +210,34 @@ country_market AS (
WHERE market_score > 0
GROUP BY country_code
),
-- Step 3: add opportunity_score using country market validation signal.
-- Step 3: country-level supply saturation — venues per 100K at the country level.
-- Used to dampen supply deficit in saturated markets (Spain, Sweden).
country_supply AS (
SELECT
country_code,
SUM(city_padel_venue_count) AS country_venues,
SUM(population) AS country_pop,
CASE WHEN SUM(population) > 0
THEN SUM(city_padel_venue_count) * 100000.0 / SUM(population)
ELSE 0
END AS venues_per_100k
FROM foundation.dim_cities
WHERE population > 0
GROUP BY country_code
),
-- Step 4: add opportunity_score using country market validation + supply saturation.
scored AS (
SELECT ms.*,
-- ── Opportunity Score (Marktpotenzial-Score v6, H3 catchment) ──────────
-- ── Opportunity Score (Marktpotenzial-Score v7, H3 catchment) ──────────
ROUND(
-- Addressable market (20 pts): log-scaled catchment population, ceiling 500K
20.0 * LEAST(1.0, LN(GREATEST(catchment_population, 1)) / LN(500000))
-- Economic power (15 pts): income PPS normalised to 35,000
+ 15.0 * LEAST(1.0, COALESCE(median_income_pps, 15000) / 35000.0)
-- Supply deficit (40 pts): max of density gap and distance gap.
-- Ceiling 5/100k (down from 8): Spain at 6-16/100k now hits zero-gap.
-- Dampened by country-level supply saturation:
-- Spain (7.4/100k) → dampener 0.30 → 12 pts max
-- Germany (0.24/100k) → dampener 0.98 → ~39 pts max
+ 40.0 * GREATEST(
-- density-based gap (H3 catchment): 0 courts = 1.0, 5/100k = 0.0
GREATEST(0.0, 1.0 - COALESCE(
@@ -230,6 +248,8 @@ scored AS (
-- distance-based gap: 30km+ = 1.0, 0km = 0.0; NULL = 0.5
COALESCE(LEAST(1.0, nearest_padel_court_km / 30.0), 0.5)
)
-- Country supply dampener: floor 0.3 so deficit never fully vanishes
* GREATEST(0.3, 1.0 - COALESCE(cs.venues_per_100k, 0.0) / 10.0)
-- Sports culture (10 pts): tennis density as racquet-sport adoption proxy.
-- Ceiling 50 courts within 25km. Harmless when tennis data is zero (contributes 0).
+ 10.0 * LEAST(1.0, COALESCE(tennis_courts_within_25km, 0) / 50.0)
@@ -247,6 +267,7 @@ scored AS (
, 1) AS opportunity_score
FROM market_scored ms
LEFT JOIN country_market cm ON ms.country_code = cm.country_code
LEFT JOIN country_supply cs ON ms.country_code = cs.country_code
)
SELECT
s.geoname_id,