What the tool estimates
The property checker is a decision-support demonstration, not an agent valuation or financial-advice service. It combines a supported purchase-price basis, local settled-sale evidence, rent evidence and explicit financing and ownership assumptions. The result is a weaker, expected and stronger buy-hold-sell scenario for 3, 5, 7 or 10 years.
Estimated profit after costs includes projected sale proceeds and rent, then subtracts the purchase price, NSW transfer duty, loan interest, property running costs and selling costs. Personal tax, renovations and loan fees are excluded and disclosed beside the result.
Evidence and geographic matching
The current snapshot covers 355 Greater Sydney locality and postcode markets. A six-month settled-sale median is used to reduce the influence of a single unusual transaction. A normal recommendation requires at least 20 recent settled sales. When an exact locality is unavailable, the interface labels a postcode proxy instead of presenting it as exact local evidence.
A public listing URL may provide visible property facts or an advertised price. The system does not sign in, bypass access controls or invent a withheld price. A suburb benchmark is always labelled as market context—not the seller’s asking price and not a valuation of the specific dwelling.
Observed historical price-error bands
Error is measured on time-based holdout data, so later sales are tested against models trained only on earlier information. Median absolute percentage error describes the typical historical difference between an estimated and actual sale price; an individual property can be wrong by more.
| Supported price range | Observed median price error | Public status |
|---|---|---|
| $350,000–$749,999 | About 20.5% | Supported with caution |
| $750,000–$1.49M | About 12.5% | Supported |
| $1.5M–below $3M | About 16.1% | Supported with caution |
| $3M+ | About 26.9% | No public recommendation |
These figures measure historical price-estimation error, not future profit accuracy. Interest rates, rents, costs and market regimes create additional uncertainty in an investment outcome.
Testing and model governance
Candidate models are evaluated through chronological walk-forward tests, unseen locations, unseen properties and stressed conditions such as missing property attributes, stale comparables and premium price segments. Interpretable baselines must be beaten before a more complex challenger can be promoted.
The research programme includes comparable-sale retrieval, regularised hedonic models, tree ensembles, gradient boosting, spatial and temporal features, conformal uncertainty and champion–challenger controls. The public calculator uses the governed scenario engine; experimental challengers are not silently placed behind the website.
How to use the result responsibly
- Compare the expected and weaker-growth outcomes, not only the headline estimate.
- Check the number of recent settled sales and whether the location is an exact match or postcode proxy.
- Replace generic assumptions with professional lending, tax, building, strata and legal advice.
- Verify the property’s condition, land, title, planning controls and seller price independently.
Open the Sydney property investment calculator and run a research scenario →