functions
01
The multiply-everything model
Independent filters are multiplied (joint probability under independence). A +12% correlation uplift stops “tall + rich” stacks from being astronomically harsher than a naive model.
Receipts & assumptions
Methodology
Delulu Dating multiplies independent population filters. It's a Fermi estimate for dating — not a prediction about any individual human.
sciencePublic stats only — no crystal ball
height
02
Height (government curve)
Census doesn't publish “mean male height” every year. We use the Population Health Survey (2014/15) male mean (169.5 cm) plus a literature SD (~5.8 cm) for the bell curve. Newer self-reported app polls inflate height — we keep the boring government curve until CHP publishes a replacement bundle.
payments
03
Income percentiles
Census and Statistics Department Annual Earnings and Hours Survey — male wage percentile curve (HK$), interpolated between published points.
cake
04
Age, papers & degrees
2021 Census and thematic reports (simplified age-window and never-married proxies).
smoke_free
05
Smoke & kids
Broad CHP / survey priors — rounded, not bespoke matchmaking intel.
apartment
06
Flat & car (HK edition)
Blunt priors on common deal-breakers. Must own flat keeps about 11% of modeled men. Must keep a car keeps about 24%. Not estate agency research.
public
07
International / local slice
~5.5% of the modeled pool for “international / non-local-raised” backgrounds — in the ballpark of Hong Kong's non-Chinese and foreign-born share (Census 2021 ethnic minorities ~8%, deflated because that headline covers all ages and our pool is already narrowed). Not passport policing.
Read this twice
This tool is a playful model built from public statistics. Use it as perspective, not fate.