PROPERTYALPHA · METHOD 2.0
Transparent value. Evidence-led risks.
Value, price advantage, net return and uncertainty are separate measures. They cannot be replaced by one historical score.
1. Comparable value
Source-linked observations stay in separate asking-price and user-reported sale groups. Same project, different known units, matching currency, positive area/price and observations within 180 days are required. Subject identity must be known. At least three eligible units within ±30% of subject area are needed. A price/m² spread greater than 50% of the median withholds the result. These are conservative workflow thresholds, not industry standards or calibrated accuracy guarantees.
Reference = median comparable price/m² × subject area. The displayed range is sample min–max, not a statistical confidence interval. Independent review of condition, floor, view, tenure, occupancy and source independence remains required.
2. Price advantage is not profit
Price advantage = (reference − recorded asking price) / reference × 100. Negative values mean asking above reference. Net profit also deducts taxes, legal and inspection fees, renovation, liabilities, financing, holding, selling costs and reserves. Regional medians are context only, never the unit resale price.
3. Risk and uncertainty
Unknown documents do not establish a defect or a 100% risk. Record issues separately for ownership/title, possession, physical condition, rental permissions, building/location, costs and exit liquidity, with sources, dates, impact and mitigation. Critical irreversible loss (demolition, expropriation or dismantling) needs an authoritative document for the exact asset and operator review of its final effect. An allegation, preliminary notice, user checkbox or high historical score is insufficient. Certainty of an event is not the percentage of money lost; compensation must be assessed separately. No calibrated probability is available in this release.
4. Practices reviewed
Reviewed Zillow and Redfin for estimates with accuracy limitations; HouseCanary for comparable selection and adjustments; PropStream for source and investment workflows; DealCheck for complete expenses; RICS for comparable evidence; MISMO for AVM accuracy/confidence testing. This release implements an explainable comparable calculation, not their proprietary models or datasets. It has not been certified by these organisations.
Next validation step: record independently evidenced sale outcomes, evaluate out-of-sample error by area and property type, then calibrate ranges. More data is needed before training or claiming a machine-learning AVM.