Models Are Pricing You Out

A new institutional-grade credit model — trained on an enormous pool of non-QM loan performance data — just went live for the securitization market. It's built specifically for the borrower profiles that traditional agency underwriting ignores: self-employed, low-doc, gig-income, small real estate investors. And it changes the competitive landscape for off-market note buyers in a meaningful way.

Non-QM RMBS issuance nearly doubled year-over-year in a single quarter in 2025, reaching a record high. That kind of institutional capital inflow doesn't arrive quietly — it compresses spreads on newly originated pools fast. The buyers who move first on this shift aren't the ones chasing new deals. They're the ones recognizing that a model trained on recent data creates a specific pricing blind spot in older loan cohorts. Those seasoned vintages — originated before the current underwriting era — are now sitting in servicer books without a clean path to securitization, and delinquency trends in those cohorts are creating both risk and opportunity depending on where you look.

This is a narrow window. Once model adoption widens across the market, the information advantage collapses. The off-market edge right now is in understanding exactly which pools get mispriced by the model — and why.

🔓 Unlock the Full Playbook

Knowing the model exists isn't enough — you need the exact mechanics of where it breaks and how to source the pools it underprices.

• The Real Numbers: The verified RMBS issuance figures, quarterly trend data from Morningstar DBRS, and the 2026 production forecast that shows how fast this market is moving.
• Where to Source: The specific vintage range to target, why delinquency trends in 2022–2024 pools create a pricing gap, and the public data tools — including a free filterable non-QM market database — that let you screen by state, county, and borrower profile before you call a servicer.
• Verification Framework: A step-by-step checklist for validating whether a seasoned non-QM pool is genuinely mispriced or correctly cheap — using real sources already tracking this market.