How accurate are home price estimates? It depends on how many similar homes have sold
We tested 130,655 Singapore home sales, estimating each one using only earlier transactions. For HDB flats and resale condos, the typical estimate missed the final price by 3.2% to 3.6%. Where a condo had no sales in the two years before, the typical miss rose to 7.6%.
Desmond Teng, founder of Reva · · 4 min read
Almost every property site will tell you what a home is worth. Far fewer show you how close their estimates come to the prices homes actually sell for.
So we tested our own. We took 130,655 sales of HDB flats, resale condos and apartments, and new launches, estimated each one using only the sales recorded before it, and compared each estimate with the price actually paid.
For HDB flats and resale condos, here’s how close our estimates came to the final sale price.
| Accuracy | Results |
|---|---|
| Within 3.5% of the final price | 1 in 2 estimates |
| Within 10% of the final price | Nearly 9 in 10 estimates |
On a $640,000 HDB flat, a 3.5% difference is about $22,000.
The results also weren't equally strong for every home. One of the clearest patterns was that estimates became less reliable when fewer similar homes had sold recently.
How recent sales affect accuracy
In a condo where 20 or more units changed hands in the two years before, the typical miss was 3.0%. In a project where nothing had sold for two years, it grew to 7.6%, more than twice as much.
Suppose you're weighing up a $1.5 million unit. In a busy development, a 3.0% miss is about $45,000. In a quiet block with no recent sales, 7.6% is about $114,000, and that is enough to change whether the deal makes sense. These are misses we saw in past sales, not a margin of error for any particular unit.
HDB flats follow the same pattern. Reva starts with flats of the same type sold in the same block over the past 18 months, and needs at least five. When the block had enough, the typical miss was 3.1%. When it had to widen to the street, it was 3.7%. When it had to draw on sales from across the town, it rose to 7.6%.
How close the estimates came
| Home | Sales tested | Typical miss | Within 10% |
|---|---|---|---|
| HDB flat | 24,573 | 3.6% | 87 in 100 |
| Condo (resale) | 40,123 | 3.2% | 91 in 100 |
| Apartment (resale) | 24,284 | 3.5% | 87 in 100 |
| New launch, once the project has sales | 20,270 | 1.9% | 96 in 100 |
| New launch, before its first sales | 21,405 | 6.8% | 69 in 100 |
What about new launches?
New launches are the hardest case, and they are where buyers commit the most money up front. Before a project records its first sales, there's nothing of its own to go on, only the launches and resales around it. The typical miss for those estimates was 6.8%.
Once a launch has sales of its own, the typical miss drops to 1.9%, and most of those estimates came after 20 or more sales. With fewer than 20, it was 3.1%.
That doesn't make an early booking a bad buy. It does mean you're leaning more on the developer's price list, and less on evidence, than a buyer who comes in later.
What this means when you're checking a price
A price estimate is more useful when you can see the evidence behind it. Before relying on a figure, from any source, ask three questions:
- How many similar homes have sold recently? If very few, treat the figure as a rough guide, not a price.
- How wide is the range? A single figure hides the uncertainty.
- Is the evidence strong enough for a confident answer? A good estimate says so when it isn't.
Reva shows the recorded sales behind its estimates and gives you a range, so you can judge how much evidence there is for the home you are considering. In our test, the final price fell inside that range about eight times in ten for resale condos, apartments and new launches. Where there are too few recent sales, Reva says so instead of guessing.
How we did it
Each estimate used only information available before the sale it was testing, as if we had been asked on the day. For resale condos, apartments and new launches, the models were retrained each year on the sales known on 1 January, each sale's inputs were limited to what was known on its own date, and the range was set from earlier years' misses. These are sales from August 2021 to August 2026 that appear in URA's records. For HDB flats, we tested 24,573 resales from September 2025 to August 2026, from HDB's records, using only the sales and the HDB price index already published when each sale took place. "Typical miss" is the median difference between the estimate and the sale price, in either direction. Download the data.
Sources: contains information from Resale Flat Prices (HDB), accessed on 19 September 2026 from data.gov.sg, made available under the terms of the Singapore Open Data Licence version 1.0; URA private residential transactions. Reva gives information, not advice, and is not a licensed estate agent or financial adviser.