Skip to content

Latest commit

 

History

History
60 lines (50 loc) · 2.88 KB

File metadata and controls

60 lines (50 loc) · 2.88 KB

Sample trace: Wallingford bungalow

A real run of the deterministic engine on the "Wallingford bungalow" demo subject, as-of 2015-05-01 (inside the King County data window). Every figure below comes from the core; the rationale is the deterministic template (no API key needed). The numbers reproduce exactly because the pipeline is pure and seeded.

Subject

Field Value
Property type detached
Beds 3
Baths 2
Living area 1,800 sqft
Lot 4,000 sqft
Year built 1960
Grade 7
Condition 3
Location 47.6795, -122.346
As-of date 2015-05-01

Deterministic output

Field Value
Conservative value (headline) $563,019
Point estimate $654,849
Range (P25 to P75) $633,214 to $715,664
Confidence Medium
Comps retrieved 20
Comps included 17
Comps excluded 3 ($/sqft outliers)
Mean comp distance 1.2 km
Median sale age 59 days
Price dispersion 12%
Mean hedonic adjustment 9%
Compute time ~30 ms

Rationale

Conservative value $563,019 is positioned below the point estimate of $654,849 based on 17 comparable sales within 1.2 km, with a median sale age of 59 days. The margin reflects 12% price dispersion among the included comparables and a 9% mean hedonic adjustment, indicating manageable market variability. Excluded 3 of 20 comparables: 3 $/sqft outlier.

With a GEMINI_API_KEY set, the same numbers are sent to Gemini Flash, which replaces this template with an interpretive lender rationale. The figures never change; only the prose does.

What this shows

  • The headline is the conservative value ($563,019), below the point estimate. A lender sizes against this, not the point.
  • 3 of 20 retrieved comps were flagged as $/sqft outliers and excluded from the math, but stay visible in the table with their reason.
  • Confidence is Medium: 17 comps clears the count and distance bars, but the price dispersion (0.1235) edges just past the High cap of 0.12, so it lands in the Medium band rather than High. It displays as 12% rounded, which is why the rounded figure alone does not tell the whole story.
  • The conservative margin reflects dispersion, distance, staleness, and the mean adjustment, capped at 25%.