flatlas.sg

How these figures are computed

Data to Sep 2026 · updated monthly

Every price on flatlas.sg is computed from recorded HDB transactions: a median of real sales, not an asking price, an average or an official valuation. This page sets out where the data comes from and exactly how each figure is made.

Data sources

All data.gov.sg data is used under the Singapore Open Data Licence v1.0.

Which months are included

Transactions are registered with a lag of one to two months, so the newest month in the feed is usually incomplete. A month counts as complete once its volume reaches 75% of the average of the three months before it. Until then, town and flat-type figures and every "as of" date stop at the month before; block medians and the map already include the sales registered so far. The latest complete months are Sep 2026 for resale and Sep 2026 for rentals.

How each figure is calculated

Today's dollars and prices as paid

Prices on these pages are nominal — what was paid. The map can also show them in today's dollars: real price = price paid × (latest CPI ÷ CPI in the year of sale), using the annual average of CPI All Items. The general CPI is used rather than HDB's own resale price index, because deflating HDB prices by HDB prices would remove the market's growth from the very figures meant to show it.

Resale prices before March 2012 are recorded by HDB's approval date; from March 2012 by registration date.

The price model

The fair-value estimates — the overpriced check and each block page's "what the price model says" — come from a hedonic regression: the log of the resale price modelled on the flat's town, flat type, flat model, floor area, storey, remaining lease, distance to the nearest MRT station and to the city centre, the plot ratio and land use around the block, and the month of sale. It is refitted every month on the 154,330 resales since Jan 2021 with complete records (a located block, floor area and remaining lease).

Accuracy is measured on a fifth of those sales held back from fitting: the model explains 94.3% of the variation in price (R²), its average error is 5.6%, and 85% of its estimates land within 10% of the actual price. The range shown around an estimate is that typical error. An estimate describes a typical flat in the block, not a specific unit, and is not an official valuation.

What these figures are not

Spotted something wrong? Use "Send feedback" in the map's About panel (ⓘ), or "Something wrong with this block?" at the foot of any block's panel.