What is your flat really worth?

An instant, data-driven estimate for any of the ~2,800 flats at Nan Fung Sun Chuen, Quarry Bay — built on 4,862 actual transactions since 1993.

—Estate index (1999 = 100)
—Typical HK$ per sq ft (saleable)
4.1%Average estimate error (MAPE)
12Blocks · 8 flats per floor
33 yrsOf transaction history
01 · Valuation calculator

Estimate a flat in three steps

Choose a block, floor and flat. We fill in its size, orientation and view automatically — you can adjust anything you know better.

1Locate the flat
Flat

2Check its features
Advanced inputs

Defaults come from the estate floor plans and Centa-WindowView. Change them only if the flat has been altered (e.g. two flats combined).


3Valuation date & price check
COMPARE · Two flats

Two homes. One clear comparison.

Choose two flats anywhere in Nan Fung Sun Chuen to compare their estimated prices — and see which modelled features explain the difference.

Preparing the comparison…
02 · Estate explorer

Every flat, in three dimensions

Drag to orbit, scroll or pinch to zoom, click any flat to see its view and value. Colour the towers by price, view, orientation or size.

Keep the original flat-level explorer, or see it amid real surrounding streets and mapped buildings.
N
3D view unavailable on this device.
Use the floor-by-floor grid on the right — it shows the same information.
BlockTap a cell to select

Schematic massing for orientation and comparison — block positions and tower shapes are simplified and not to scale. Flats A/H face north, B/C east, D/E south, F/G west.

03 · Estate price index

33 years of quality-adjusted prices

Built from the monthly time dummies of the hedonic model, so it tracks price changes for a flat of constant quality — not just the mix of flats that happened to sell.

Hollow points = months with no qualifying sales (interpolated).
Latest—
Change in selected range—
Peak in range—
Annual averages
04 · Implicit prices

What makes one flat dearer than another?

Each chart shows the model's estimate of how much one feature changes price, holding everything else equal. Grey badges mark effects that are not statistically reliable.

05 · The estate

Nan Fung Sun Chuen at a glance

Blocks of Nan Fung Sun Chuen seen from the estate podium, Quarry Bay.
06 · Methodology

How the estimate is made

A hedonic price model: each flat's price is explained by its features and the month it sold.

Final model (log-linear, monthly time dummies)

ln(NP) = α₀ + α₁SFA + α₂SFA² + α₃AGE + α₄AGE² + α₅FL + α₆FL² + α₇LUCKY + α₈DMTR + α₉DCARPARK + α₁₀DMARKET + α₁₁BED + Σ ORIENTATION + Σ VIEW + Σ γtTt + AR(1)
R²0.987Explains 98.7% of price variation
MAPE4.11%Average gap, estimate vs actual
Durbin–Watson2.02AR(1) term added — no serial correlation
Breusch–Pagan–Godfreyp < 0.01Heteroskedastic → White-robust errors; coefficients unchanged

Data

  • EPRC transaction records, Aug 1993 – Jun 2026: 5,274 records, 4,862 used after cleaning.
  • Removed: sales with car-parking spaces, $0 or duplicate records, uncompleted / pre-sale deals, and outliers with residuals beyond ±0.4.
  • Added attributes: GeoInfo Map distances, Centaline floor plans (orientation, bedrooms) and Centa-WindowView (views).

Estimated coefficients

ARMA GLS with AR(1); 378 monthly dummies (base: Jul 1993) not shown — see the Price Index section.

Limitations users should know

    07 · Updating & maintenance

    Keeping the model honest

    House prices move every month. This is how the tool stays current without silently rewriting history.

    Quality gates

      Who does what

        Version log

        08 · Help

        User guide & FAQ

        Help us improve the estimate

        Know a recent sale or think the estimate is off? Tell us — feedback is reviewed at each model update.