Percentile derivation. Each verdict card's percentile
(e.g. "72nd percentile") ranks this parcel's score against a Toronto
reference sample (n=138 parcels) drawn from the training-snapshot
substrate. A 72nd percentile means your parcel scores better/higher than
72% of the sampled Toronto parcels for that metric.
Band thresholds. Each arrow assigns one of five
bands based on percentile + domain-specific rules:
LOW (best-case, typically bottom-quartile risk or
top-quartile amenity), MED, HIGH,
EXTREME (worst-case for risk arrows, best-case for amenity
arrows), or NO_DATA (data unavailable for this parcel;
excluded from composite).
Composite score. Verdicts contribute to a weighted
composite ranked into two axes: composite_risk
(higher = worse) and composite_livability
(higher = better). Weights reflect empirical impact-on-decision from
shipped case data, not equal-weighted averages. Individual arrow
weights are visible in each card's provenance footer.
Data sources. NRCan flood history, TRCA regulatory
floodline (GTHA parcels), Health Canada Cross-Canada Radon Survey
(FSA regional averages), Ontario Ministry of Education EQAO scores,
NASA FIRMS + USGS + NOAA (wildfire / earthquake / smoke), City of
Toronto Open Data (permits, zoning, hydrants, fire stations), Nominatim
(geocoding), TTC + Metrolinx GTFS (transit).
Limitations. Reference sample sizes are noted in
each card. Data ingest cadence varies by source (typically weekly for
permits / zoning; monthly for wildfire; annual for EQAO). Refer to your
RECO agent + FSRA broker + qualified home inspector for
transaction-critical verification.