Risk score
Risk score
Section titled “Risk score”Risk is scored separately from opportunity and is never blended in. A deal can read high-opportunity and high-risk at the same time — that is a valid, visible state, not a contradiction. Collapsing the two into one number is how tax-trap yields end up looking like deals.
How it works
Section titled “How it works”Each triggered flag carries a stated reason and a severity contribution. Score = sum of severities, capped at 100. Bands: LOW < 12, MODERATE 12–29, ELEVATED 30–54, HIGH ≥ 55. Returned on every underwrite and via the risk_report MCP tool as {score, band, flags[]}.
| Flag | Severity | Trigger (model p5v4-2026.07) |
|---|---|---|
| Tax shock | +22 / +10 | township’s last triennial reassessment moved median AV ≥ +30% / ≥ +15% |
| Legal-unit mismatch | +20 | listing claims more units than the assessor class implies — financing/insurance/resale blocker until cured |
| Basement flood | +18 / +8 | ≥ 80 / ≥ 25 water-in-basement 311 calls (3 yr) nearby |
| Above-neighborhood ARV | +16 | target ARV > 110% of the 75th-percentile comp — appraisal ceiling |
| Tax trap | +16 | effective rate ≥ 5% |
| Thin comps / appraisal risk | +14 | comp confidence < 40 with < 4 usable comps |
| FEMA flood zone | +12 | in/near a special flood hazard area, or flood factor ≥ 7 |
| Distress proximity | +12 | < 1.5 mi from a financially distressed municipality |
| Reassessment imminent | +8 | township reassesses this year or next |
| Closing friction | +6 | < 1.2 mi from a point-of-sale-inspection town |
| Vintage stock | +6 | built pre-1940 — lead/asbestos protocol, service upgrades, insurance friction |
| Rent-model uncertainty | +8 | the modeled rent evidence band is wider than the supported threshold |
Coverage and display caveats
Section titled “Coverage and display caveats”The basement-311 flag is Chicago-only. Outside Chicago that field is not tracked, not zero, and fewer than three applicable signals produce an UNKNOWN band rather than a reassuring LOW. Crime and protected/proxy demographic fields are excluded from public property risk and recommendation outputs (fair housing). Severities are hand-set and additive; they have not yet been calibrated against realized losses — treat the band as a checklist prioritizer, not a probability.