5035 Underwood

Method and limits. MSI v0.2-detroit-2026-09 weights twenty public signals by how much each one changed the recorded three-month arm’s-length sale rate across all 214,000 taxable residential-improved parcels in the City of Detroit, using the Assessor’s February-2026 roll so every owner-side signal predates the sales it is measured against, checked on a holdout the model never saw (the top 0.5% of homes by score sold at 5.9x the citywide rate, the top 1% at 5.7x).: reboard / vacant-property / squatter request (improve detroit) +17, mechanical / hvac permit +16, two or more blight tickets +14, bought since 2024 +10, weeds / vegetation blight ticket +10, taxpayer mails out of state +9, plumbing permit +5, rodent blight ticket +3, dumping / debris blight ticket +2, two or more units -4, owned for 20 years or more -6, neighborhood enterprise zone parcel -8, blight fines still due -10, failure-to-maintain blight ticket -11, tall-grass / weeds request (improve detroit) -14, electrical permit -16, inoperable-vehicle blight ticket -16, no recorded sale -16, assessed under $5,000 -16, principal residence exemption (owner-occupied) -20. The "in 100 homes" figure is the model's fitted probability for this combination of signals. It describes homes like this one in aggregate; it is not a claim about this owner's intentions, and it is not LienSuite's paid deal score.