Skip to content
Back to Alex Score
Alex ScoreMethod v1.0 · Method date 2026-10-06

A rating you can take apart.

Alex Score turns explicit inputs into an explainable comparison. This page publishes the rules used by the scoring engine, including the limits of what a number can tell you.

Category-specificA mortgage score and a card score measure different things.
Rules, not a predictionWeights are MFA editorial choices, not statistically validated probabilities.
Independent of commissionsCompensation, popularity and sponsorship are not scoring inputs.

What Alex Score means

Alex Score is a 0–100 educational index of modeled value and fit under the entered assumptions. It rates a scenario, not the overall quality of a lender, issuer, bank or person. Different inputs, targets, horizons or comparison sets can produce different scores for the same offer.

These formulas and weights are MyFriendAlex’s original, deterministic heuristics. They have not been empirically validated as predictors of financial outcomes, borrower approval, creditworthiness or product safety. A score of 90 is not a 90% chance of success.

Component points = weight × clamp(component value, 0, 100) ÷ 100.
Overall score = round(sum of unrounded component points), after any stated cap.
Clamp keeps a component between 0 and 100. All weights in each category sum to 100.
90–100Strong modeled fit
75–89Worth comparing
50–74Meaningful trade-offs

0–49: weak modeled fit. When a stated payment, upfront-cost or admin-time limit is exceeded, the label becomes Outside your limits and the maximum displayed score is 49. Missing or unsupported inputs produce no score and a list of what is needed. Equal modeled dollars are explicitly treated as a tie; scores are not tie-breakers for qualitative factors.

Compare scores within one category using consistent assumptions. Always inspect the dollar figures: small score differences can accompany substantial costs. A strong score against an easy target does not establish that an offer is competitive in the wider market.

Mortgage score: the same loan need, two quotes

Version 1 supports exactly two fixed-rate, fully amortizing loans with the same principal. Loan terms may differ. Enter a shared holding period, a monthly principal-and-interest plus PMI limit, and an upfront loan-cost limit. The comparison does not assess underwriting or total home affordability.

First calculate borrowing cost

Monthly P&I = P × r ÷ [1 − (1 + r)−n]. For a 0% loan, use P ÷ n.
P = shared principal; r = annual rate ÷ 1,200; n = loan term in months.
Holding months = round(entered years × 12).
Borrowing cost = amortized interest + constant monthly PMI over the modeled loan months + upfront loan costs − lender credits.

The engine steps through monthly amortization. Principal repayment is shown as a balance reduction, not charged again as borrowing cost. Costs stop at payoff if the loan matures before the horizon. Enter loan costs including points, without also adding points a second time. Excess lender credits are unsupported: if credits exceed the entered loan costs, no score is assigned.

Then award points

60 points — Borrowing-cost value: component value = 100 × lower modeled borrowing cost of the two quotes ÷ this quote’s borrowing cost. A zero-cost quote receives 100 for this component; if only one is zero, the nonzero-cost quote receives 0.
25 points — Payment-limit fit: 100 × min(1, your monthly P&I + PMI limit ÷ this quote’s monthly P&I + PMI).
15 points — Upfront-cost fit: 100 × min(1, your upfront loan-cost limit ÷ this quote’s net upfront loan costs). Zero net costs receive 100.

Exceeding either monetary limit by more than half a cent caps the overall score at 49. The limits must be explicitly entered; an upfront-cost limit of 0 is valid. Different loan terms also change monthly principal repayment and cash flow, which you should review alongside borrowing cost.

Excluded: down payment, property taxes, homeowners insurance, HOA, escrows, maintenance, tax effects, investment opportunity cost, refinancing, extra payments, balloon payments, prepayment penalties, lender service and closing reliability. ARMs and non-amortizing loans are unsupported. PMI remains constant in the model; it does not infer cancellation.

The lower-cost quote receives full cost-component points relative to the other entered quote. Both quotes could still be expensive compared with the market. Compare matching quote dates and equivalent terms before acting.

Credit-card score: recurring value after costs

Enter estimated annual rewards value, annual perks you would actually use, annual card and incremental membership fees, estimated annual interest, and a shared positive annual net-value target. The target is your chosen comparison threshold, not an MFA promise or market benchmark.

Gross value = annual rewards value + usable annual perks.
Net annual value = gross value − annual fees − estimated annual interest.
80 points — Net value vs target: component value = clamp(100 × net annual value ÷ your target, 0, 100).
20 points — Value retained after fees: component value = 100 × max(0, gross value − annual fees) ÷ gross value. If gross value is 0, use 0.

No positive net annual value means an overall score of 0, including the fee-efficiency component. Negative modeled dollars remain visible; they are not replaced by a $0 net-value claim. The fee-efficiency component measures the share of benefits retained after fees, before interest. Interest affects the net-value component and the nonpositive-value gate.

Entering 0 interest assumes no interest cost. If you carry balances, use a realistic annual interest estimate or the payoff calculator. A card comparison handoff deliberately leaves interest blank until you supply it. MFA does not infer it from your rewards spending.

Excluded: welcome bonuses, approval odds, credit-score impact, unvalued protections, changing point prices, category caps unless already included in the entered rewards estimate, redemption effort and merchant coding. Variable point values are assumptions, not guaranteed cash. Usable perks should reflect value you would obtain without unnecessary spending.

Card-comparison handoffs use the existing category-aware rewards model, selected point assumptions, entered usable perks and incremental fee calculation. Alex Score does not replace that calculation or change the existing card ranking.

Bank-bonus score: net dollars, liquidity and effort

Version 1 covers cash bank bonuses only. Enter the cash bonus, all fees over the period, other incremental costs, cash tied up, discretionary cash after emergency reserves, holding days, offer and alternative APYs, admin hours, your acceptable hours and a positive net bonus target.

Foregone interest = max(0, tied-up cash × [(1 + alternative APY ÷ 100)days ÷ 365 − (1 + offer APY ÷ 100)days ÷ 365]).
Net bonus before tax = cash bonus − fees − other incremental costs − foregone interest.

APYs are held constant. This uses the difference in modeled compound yields over the entered period. Higher offer yield does not increase the bonus figure; foregone interest simply stops at 0. Money is assumed tied up for the whole entered period. Holding periods and fees are not inferred from incomplete marketing text.

70 points — Net bonus vs target: component value = clamp(100 × net bonus ÷ target, 0, 100).
20 points — Discretionary-cash headroom: 100 × (1 − cash tied up ÷ discretionary cash). No tied-up cash receives 100.
10 points — Admin-time fit: 100 × min(1, acceptable admin hours ÷ estimated admin hours). Zero estimated hours receive 100.

Gates: no score if eligibility is unconfirmed or not met, or if required cash exceeds available discretionary cash. Selecting Yes is your assertion that the offer is current and geography, direct-deposit cadence, new-customer conditions and other requirements are met. No positive net bonus means an overall score of 0. Admin hours above your limit cap a positive-value score at 49.

Excluded: taxes, effort valued in dollars, offer payout risk, deposit insurance analysis, approval, clawback probability, changing rates and card welcome bonuses. Confirm actual terms and any account-open duration, post-payout holding requirement or fee waiver with the institution.

Data confidence is separate from the rating

Low: sample figures, unreviewed edits, linked inputs, or restored browser scenarios. An edited scenario can still contain unchanged sample values. Missing required numbers stay missing rather than becoming zero.

Medium: you explicitly mark the inputs and current terms reviewed. This is a self-reported check. MFA has not independently validated the document, issuer terms or your eligibility. Editing any input removes the review marker; reopening a browser-saved scenario also requires another review.

Version 1 does not assign a High confidence label. Complete mathematical inputs do not prove reliable source data. Calculation and snapshot dates are never presented as source-verification dates.

Versioning, independence and saved results

Version 1.0 · 2026-10-06: initial release of the mortgage, recurring-card and cash-bank-bonus models above. No machine learning, popularity data, institution names, affiliate compensation or sponsorship affects the score.

Formula, weighting, gate or label changes require a new scoring version and updated methodology. Saved score reports and Decision Lab captures retain the version and their calculated snapshot. Reopening the tool applies the current method to the current or browser-saved inputs; it does not silently rewrite an old saved score.

Inputs calculate in your browser. Browser scenario saving is explicit and can be removed with Forget saved scenario. Downloaded reports include entered numbers. Keep in Decision Lab queues a snapshot for you to attach and save; it does not automatically publish or create a public share link. These numbers are not appended to the page URL during editing.

Alex Score does not alter the existing quarterly readiness or credit-card reward models. Readiness, decision-completeness and other categories have no Alex Score in version 1.

Insights, stress tests, history and quote import

Feature release · October 6, 2026: these tools use scoring method v1.0 without changing its weights, formulas or gates.

Alex Insights and dollar thresholds

Mortgage cost-leader changes are checked monthly from month 0 through month 480. A reported reversal is the first month on that grid with a different non-tied cost leader; it is not an exact daily break-even date. A difference below half a cent is treated as a tie. Fees-to-tie holds rate, term, principal and PMI fixed: required net upfront costs = this quote’s net upfront costs + the other quote’s borrowing cost − this quote’s borrowing cost. If that result is negative, removing fees alone cannot close the gap. Payment and cash limits remain separate gates.

Card interest break-even equals annual rewards plus usable perks minus annual fees. At that annual interest amount, net value is zero. Target and comparison gaps are differences in modeled net dollars; they are not instructions to spend more or invent perk value. Bank-bonus insights separate fees, other costs, foregone interest, cash headroom and the stated admin-time limit. Taxes and effort valued in dollars remain outside the model.

Stress scenarios are hypothetical

Mortgage scenarios use 1, 3, 5, 10 and 20-year holding periods, keeping both fixed rates, terms, upfront costs, PMI and personal limits unchanged. Card scenarios reduce entered rewards by 0%, 10%, 25%, 50% and 75%; usable perks and fees stay fixed. The interest slider adds the same dollar shock to each card’s entered annual interest. Bonus scenarios add 0, 30, 60, 90 and 180 holding days; the APY slider adds percentage points to the alternative APY, leaving offer APY and the other inputs unchanged.

Every scenario reruns the original engine with all its limits. Missing inputs remain missing. Values outside the engine’s supported range produce no score. Charts and tables compare borrowing cost for mortgages and net value for cards/bonuses, separately from score. No likelihood or forecast is assigned, and stress controls do not edit or save the baseline.

Named snapshots and score cards

Score history stores up to 20 explicit snapshots across the categories in this browser. Each retains its entered inputs, calculated results, self-reported confidence, timestamp and method version. Old results are displayed as saved. Restoring loads old inputs under the current method and clears the review marker; it does not overwrite the old snapshot. Snapshot-to-current comparisons list the changed inputs and the total score/dollar difference, without claiming to allocate the effect of interacting inputs to individual factors. Comparisons across method versions are labeled accordingly.

Snapshots can be individually downloaded as JSON or removed. A downloaded snapshot can be restored in another browser through a file preview, with a 256 KB limit. Restore supports the current method version only and checks that the saved scores, components and modeled dollars agree with the stored inputs within floating-point tolerance. This calculation consistency check does not authenticate the file or verify offer terms. Import is explicit, preserves the original saved results and timestamp, adds a separate import date, prevents duplicate identifiers and clears the input-review marker on loading. No file is uploaded, and a failed save leaves current inputs unchanged. Unsupported files remain available outside the app. They are separate from the single saved browser scenario and from private account records. A downloaded PNG contains the score, component points, confidence, category, method version and generation date. Dollar amounts and the main monetary assumption are hidden by default and appear only when you opt in. Example inputs are prominently labeled. Creating a score card does not upload it or create a public sharing URL. A generation date is not source verification.

Loan Estimate PDFs: suggestions before confirmation

The importer processes text-based Loan Estimate PDFs locally in the browser, up to 10 MB and 12 pages. For standard two-column Loan Estimate layouts, it looks for labeled loan amount, interest rate rather than APR, loan term, Section D total loan costs, lender credits and monthly mortgage insurance. The first payment column may suggest PMI; you must confirm the chosen constant modeled amount rather than assuming automatic cancellation. Suggestions retain their source page and text location, highlighted on a rendered document preview. Extraction does not authenticate the document or verify current terms.

Each numeric value must be reviewed and confirmed by you, including any entered zero. Missing, ambiguous or unsupported values stay blank. You must also confirm the fixed-rate, fully amortizing product assumptions and the constant modeled PMI amount. Detected ARMs, interest-only features, balloons or prepayment penalties block application. Password-protected and image-only PDFs need an unlocked text-based copy or manual entry; this release does not perform OCR.

Applying an initial PDF clears example quote values, the sample holding period and example payment/upfront limits. Your horizon and limits must be entered before scoring. A second quote must use the same principal. Lender credits above loan costs are unsupported. Applying confirmed values clears the general review marker and does not raise data confidence automatically. PDF bytes, filenames and extracted text are not uploaded or saved in history; confirmed numeric inputs become browser drafts and may also be explicitly saved as private account snapshots. Quote metadata you add is kept with the saved comparison. Clear PDF and suggestions removes the document from the active import flow. Leaving the page clears it.

Keeping, revisiting and sharing decisions

Alex Score drafts automatically stay in this browser after you change the example. Optional account snapshots retain original inputs, results, confidence and method version privately in My MFA. Reopening a saved result loads those exact inputs and shows its original result separately from the current calculation. Review current offer terms again.

Mortgage quote names, quote dates, expiry dates and source notes are user supplied metadata. They do not verify a lender, offer, quote or rate lock. Expired dates prevent marking the comparison currently reviewed. Downloaded reminders require calendar import and do not update previously imported events.

Decision outcomes compare user recorded estimated and actual amounts using the same measure and period. They are not verified savings or evidence of causation. Household workspaces share only decision copies explicitly selected by the user, including captured inputs and quote notes. Viewer and editor roles are enforced by the server. Every shared save retains a version; members can review parallel edits. Membership revocation blocks future access but cannot recall downloaded copies.

Primary educational references

These sources inform what costs and trade-offs to inspect. The agencies do not endorse Alex Score, its weights, bands or formulas.

Educational estimates only. Results depend on what you enter. Use current offer documents and professional advice where appropriate before making a financial commitment.