We price every flat before it sells. Then the sale happens - and we show you both numbers.
Anyone can claim their estimates are accurate. This page removes the claiming: every HDB resale in Singapore, our call next to the hammer price - hits, misses, all of it. Over the past 12 months that was 23,704 sales, and half our calls landed within 6% of the real price.
“under” = it sold for more than we said; “over” = less. Every call made by a model that had never seen that month’s sales.
Steady, month after month
Out of every 10 flats we price, about 7 sell within 10% of our number - on a $600k flat, that means within $60k. Each point below is one month’s hit rate. The line barely moves: this isn’t a lucky streak, it’s how the model performs every month.
each column = one month · filled dots = calls within 10% · exact figures in the table
The record, month by month
| Month | Sales checked | Typical miss | Within 10% |
|---|---|---|---|
| 2026-07 | 2,081 | 6.2% | 72% |
| 2026-06 | 2,129 | 6.4% | 70% |
| 2026-05 | 2,128 | 6.5% | 70% |
| 2026-04 | 1,937 | 6.4% | 71% |
| 2026-03 | 2,044 | 6.9% | 68% |
| 2026-02 | 1,662 | 6.3% | 72% |
What pricing 235,828 flats taught us
To call prices this precisely, the model had to learn what each thing is worth. These aren’t opinions - they’re what the market actually paid, measured.
On a $600k flat, one year of lease costs more the less you have left - the slide steepens. Exactly what theory predicted; now measured.
≈ $4,760 per floor on a $600k flat, everything else identical. Where a town sits matters far more:
Folklore says living within 1km of a hot primary school adds 5-15%. We checked 231 schools, comparing like-for-like flats inside the ring against just outside it: the average premium is zero, and it doesn’t track ballot demand at all (correlation -0.07). For HDB flats, the ring premium is a myth.
Where you’ll meet these numbers
Every block page shows its own fair-value line - what that block’s flats should cost on fundamentals, from this same model. Find yours: browse the towns or type your postal code in the search box above.
How it works, honestly
A transparent statistical model (hedonic regression - the method used in academic housing research) learns from every recorded resale: size, floor, remaining lease, flat type, town, and the month’s market. No black boxes; every effect above is a number you can read. Each month is scored by a model that had never seen that month’s sales. Data: HDB resale transactions via data.gov.sg. Fit R² 0.91 on 235,828 sales · model 769f43a62a2d · not financial advice.