TrueTrend is how we forecast HDB resale prices. This page explains exactly how it works, how accurate it is, and what it cannot do. We would rather show our working than ask you to take a number on faith.
Every figure starts from official HDB resale transactions published on data.gov.sg. We use the full record from 2012 to 2026 Q2, several hundred thousand actual sales. For each town and flat type we calculate the median price every quarter. The median (the middle sale) is used rather than the average, so a handful of unusually high or low sales does not distort the picture. We work in quarters rather than months because monthly figures are noisier and a quarter still gives a timely read.
TrueTrend measures the compound growth rate of the median price over the most recent three years, then carries that rate forward, damped: the growth rate is faded slightly the further out we project. A raw trend tends to over-shoot, so easing it off produces more accurate forecasts. In plain terms: we look at how fast prices have actually been compounding lately, project that pace, and taper it rather than assume it runs forever. We deliberately keep it simple and transparent. There is no black box and no undisclosed "AI", because a forecast you cannot check is a forecast you cannot trust.
We chose this setup by testing, not by preference. Compared with a straight-line trend, with assuming prices simply stay flat, and with an un-damped compound trend, damped compound growth over a three-year window was consistently the most accurate across the full period. We also tested transaction-volume momentum as an extra signal and dropped it, because it showed no reliable predictive value. We add only what measurably improves accuracy.
We backtested it. Backtesting means going into the past, forecasting a date whose real outcome we already know, and measuring the error. We did this repeatedly across 2012 to 2026, a period that includes the 2013 to 2019 cooling decline, the 2020 to 2022 surge, and the current plateau, so the method is judged through rising, falling, and flat markets rather than one lucky stretch.
| Horizon | Typical error | In plain terms |
|---|---|---|
| 1 year ahead | about 3.8% | roughly 96% accurate |
| 2 years ahead | about 6.3% | roughly 94% accurate |
For context, a $630,000 median with a 3.8% typical error means the forecast is usually within about $24,000 of the eventual price one year out. That is the honest number, measured across a full market cycle, not a best-case figure from a single uptrend.
It cannot call turning points. Its largest historical miss was the 2020 to 2022 boom, which it under-predicted by around 15%, because nothing in the prior data signalled a sudden surge. The same blind spot applies to a sharp downturn. TrueTrend tells you where prices are heading if the recent trend continues. It is most reliable in stable or steadily trending markets, and least reliable at the exact moments the market changes direction. It is general information to inform your own research, not a valuation, a guarantee, or financial advice.
We publish forecasts for the overall market, for each flat type, and for individual towns where there is enough consistent data. Where a town-and-flat-type segment is too thin or too volatile to forecast responsibly, we leave it out rather than show an unreliable number. Everything is recomputed as fresh HDB data is released.
TrueTrend methodology, SG Property Index. Data source: HDB resale transactions via data.gov.sg. Last reviewed 16 July 2026. General information only, not financial advice.