Deciding between two specific structures (e.g. 15y vs 30y)? →
Net worth & break-even comparison holds your income fixed and
finds the return at which the choice flips.
Ranks every scraped mortgage by “vs par”: the net present
value of everything you would pay (points at closing + every monthly payment,
discounted at your opportunity cost of capital) minus the principal borrowed.
Negative = the loan is cheaper than your capital — borrowing
creates value. Positive = it destroys value versus paying cash or investing.
ARMs are priced under four rate paths: base (teaser never moves),
fully-indexed (resets step toward today’s index + margin
— the sort column), +150bps, and cap-max.
Caps are assumptions — set them from a loan’s disclosure. The exit /
refi knobs truncate the loan early (points are sunk when a loan dies); the tax
knobs credit deductible interest and purchase points above your standard deduction,
with the $750K acquisition-debt cap pro-rated.
Cheapest loans vs par
Institution
vs par
Rate
Points
Product
P&I / mo
Base
+150bps
Cap-max
Paid off
Updated
Cheaper, but gated
These would rank above, but fail your borrowing-state or
membership filter. Shown so a filter never silently hides a better loan.
Institution
vs par
Rate
Points
Product
Why gated
Methodology & caveats.
vs par = NPV of all outflows (points at t=0 + each monthly
payment, discounted monthly) minus principal borrowed, sorted by the
fully-indexed path. Fixed loans have one path, shown in the sort column.
Exit / refi: an exit year terminates every loan with a
discounted lump payoff of the remaining balance — modeling a sale or
refi, after which all loans converge. A refi chance %/yr probability-weights
truncations across the loan's life instead. Both exist because points are a
prepaid rate discount that is sunk the moment the loan dies: buydowns that
win on a 12-year hold can lose on a 3-year one.
Tax: with a marginal rate set, each year’s deductible
interest (pro-rated by the $750K acquisition-debt cap) plus year-one purchase
points are credited at that rate, but only above your standard deduction net
of other itemized deductions. This is the federal Schedule A shape; PA-style
flat state taxes offer no mortgage deduction.
Prepayment: the extra-payment window applies fixed extra
principal between two months; the glide recasts the payment at the chosen
year to amortize to zero by the deadline year. ARM payments recast at every
rate change, matching how servicers adjust.
Lender fees other than points are not included (compare Loan Estimates; a
wide rate-vs-APR gap signals fees). Advertised rates assume top-tier credit.
Refinance-only, home-equity, land, construction, first-time-buyer,
interest-only and special-program products are excluded.
The computation core is a line-for-line port of
tools/npv_reference.py, pinned to its golden outputs by tests
(the tax and hazard layers are additive wrappers). There is deliberately no
“recommended loan” output: the honest product is the sensitivity
structure, and the recommendation belongs to whoever owns the discount rate.
Workbook: the downloadable .xlsx re-expresses this engine as
plain spreadsheet formulas (tested against the same goldens), and each loan
carries the value this page computed at default inputs plus a self-check
cell — at default inputs it must read OK, so any drift between the
file and the site is visible in the file itself.