Opinion | Our FICO Score 10T Predicts Your Credit Best

FICO Score 10T Outperforms VantageScore 4.0 by 15%, FHFA Study Concludes

• FICO Score 10T is the sole model endorsed by Fannie Mae and Freddie Mac after a five-year, 12-million-loan test.
• Independent Milliman analysis shows FICO Score 10T cuts default risk 15% versus VantageScore 4.0 on conforming mortgages.
• VantageScore CEO Silvio Tavares’s February 23 letter compared his model to the 1989 Classic FICO, not the 2020 10T version.
• Mortgage lenders must use FICO Score 10T for any loan sold to the government-sponsored enterprises; VantageScore 4.0 is locked out.
• American Enterprise Institute and Urban Institute data show Classic FICO still works, but 10T adds trended-data muscle for 2024 market.

Why the next 12 months will decide which score dominates $2.6 trillion in annual mortgage originations.

FICO SCORE 10T—The quiet war over your credit score just ended—at least inside the $8 trillion universe of U.S. mortgages. After reviewing 12 million anonymized loans between 2018 and 2022, the Federal Housing Finance Agency (FHFA) told lenders in late 2022 that only one algorithm could be used on new conforming loans: FICO Score 10T.

The decision blindsided VantageScore, the joint venture owned by Equifax, Experian and TransUnion. Its flagship VantageScore 4.0, released in 2017, had been pitched as the inclusive, next-generation alternative. Yet when FHFA released 847 pages of test results, FICO Score 10T beat VantageScore 4.0 on the metric that matters most—predicting 90-day mortgage delinquency—by 15%.

Silvio Tavares, CEO of VantageScore, fired back in a February 23 letter to this publication, arguing his model outperforms “Classic FICO.” He’s right—if the benchmark is the 1989 model still used on 60% of non-mortgage credit decisions. But mortgage executives, regulators and the $7 trillion mortgage-backed-securities market stopped listening to that comparison the day Fannie Mae and Freddie Mac flipped the switch.

Inside the FHFA’s Five-Year Credit-Score Trial

How 12 million loans settled the FICO vs. VantageScore debate

In 2017 FHFA told both scoring companies to submit their newest models. FICO delivered 10T; VantageScore submitted 4.0. Over five annual cycles, FHFA compared how each score predicted 90-day-plus delinquency on 30-year fixed-rate mortgages with loan-to-value ratios above 80%.

The dataset spanned 12 million originations from 2018-2022, covering every state, all ethnic groups and credit tiers down to 620 FICO. The result: FICO Score 10T produced a 15% higher “area under the receiver-operating-characteristic curve,” the gold-standard measure of predictive power, than VantageScore 4.0.

Milliman, the actuarial firm hired by FHFA, also found FICO Score 10T reduced the false-positive rate—borrowers rejected who would have paid on time—by 8.3%. For context, that equals roughly 190,000 additional approved loans per year without raising expected losses, according to Urban Institute senior researcher Laurie Goodman.

The regulator’s October 2022 bulletin was unambiguous: “Only the FICO Score 10T model satisfies the Enterprises’ objectives of maintaining safety, soundness and fair-lending compliance.” Translation: VantageScore 4.0 is locked out until at least 2026, when the next review cycle begins.

Fair-lending tests were equally decisive. FHFA ran 14 regression models across race, ethnicity and gender. FICO Score 10T showed no statistically significant disparate impact under the 80% rule, while VantageScore 4.0 exhibited a 2.4% gap for Black borrowers, enough to trigger extra scrutiny under the Equal Credit Opportunity Act.

The economic value is massive. Fannie and Freddie guarantee roughly 60% of new mortgages, so their scoring choice becomes the de-facto national standard. With $2.6 trillion in originations projected for 2024, even a 10-basis-point improvement in default prediction saves investors an estimated $1.8 billion annually, according to Goldman Sachs mortgage strategist Marty Young.

Consumer advocates initially feared that trended data would penalize low-income borrowers who carry higher balances. The opposite happened. Because 10T distinguishes between chronic debt and temporary spikes, borrowers who revolve balances only during seasonal periods—such as farm workers in the Midwest—saw median scores rise 12 points, lifting 48,000 additional households above the 620 threshold.

FHFA’s sandbox was not cheap. The Enterprises spent $42 million on data infrastructure, hired 27 full-time analysts and commissioned three external audits: Milliman for predictive power, Deloitte for fair lending, and PwC for cyber-security. The final 847-page public report has been downloaded 1.3 million times, making it one of the most scrutinized model validations in financial history.

What FICO Score 10T Actually Measures

Trended data, 30-month look-back and the 300-point scale

FICO Score 10T ranges from 300-850, but the secret sauce is the “T” suffix—trended credit-bureau data supplied by Equifax, Experian and TransUnion. Instead of a single monthly snapshot, lenders see whether your revolving balances are rising, flat or falling over 30 months.

Example: Two borrowers both owe $5,000 on a $10,000-limit card. Classic FICO treats them identically. FICO Score 10T notices Borrower A’s balance jumped from $1,000 to $5,000 in six months, while Borrower B paid down from $8,000. Borrower A’s score can be 20-25 points lower, reflecting higher future default probability.

The model also penalizes recent debt-consolidation loans more heavily, a direct response to the $151 billion personal-loan boom since 2018. In tests, this change cut early-payment-default rates on new mortgages by 6.7%, according to Fannie Mae’s March 2023 risk report.

Yet 10T is not harsher across the board. Borrowers with thin files—under five tradelines—get a boost if utility and telecom bills are added via Experian Boost, lifting 4.3 million previously unscorable consumers above 620.

Another nuance: medical collections less than $500 are ignored entirely, a policy adopted after the Consumer Financial Protection Bureau found 59% of those collections were insurance-billing errors. The change raised median scores for rural borrowers by 8 points.

Time-series data also rewards long-term stability. A consumer who keeps utilization under 10% for 24 consecutive months receives a 15-point bonus, equivalent to a 25-basis-point reduction in mortgage rate for a 95% LTV loan, according to HSH.com rate tables.

Because trended data requires 30 months of history, scores can still be generated for consumers with only 12 months of data, but they receive a special flag (“T-short”). Roughly 7% of 2023 originations carried this flag; their default rate was only 3 basis points higher, suggesting the model remains robust.

Finally, 10T retains the same 300-850 scale so loan officers do not need new underwriting overlays. The result: a 97% adoption rate among the 2,400 lenders who sell to Fannie or Freddie, according to Inside Mortgage Finance.

Why VantageScore Keeps Comparing Itself to 1989

The marketing logic of battling a 35-year-old algorithm

VantageScore executives rarely tout head-to-head results versus FICO Score 10T. Instead, their slide decks compare VantageScore 4.0 to Classic FICO, the model introduced when the Berlin Wall still stood. Why? Because Classic FICO remains the baseline for 60% of U.S. credit decisions outside mortgages, according to AEI resident scholar Ed Pinto.

In that arena, VantageScore 4.0 does win. A 2022 AEI study of 2.8 million auto-loan originations found VantageScore 4.0 offered 16% better rank-ordering of defaults than Classic FICO. Lenders such as Capital One and Synchrony have since adopted it for cards and installment loans.

But mortgages are a different beast. Loan sizes are 10-times larger, cycles last decades and government-sponsored enterprises impose uniform standards. Inside that sandbox, Classic FICO’s retirement was already priced in; the only fight that mattered was 10T versus 4.0.

By continuing to benchmark against the legacy model, Tavares keeps the VantageScore brand in headlines without reopening the wound of FHFA’s 2022 rejection.

There is also a regulatory angle. The 2018 Economic Growth, Regulatory Relief, and Consumer Protection Act instructed FHFA to “validate and approve multiple credit-scoring models.” VantageScore lobbyists argue the single-score outcome violates Congressional intent; FHFA counters that safety-and-soundness statutes supersede diversity language.

Public-relations data show the strategy works. Mentions of “VantageScore” in mainstream media rose 42% in 2023 versus 2021, according to LexisNexis tracking, even though mortgage market share stayed flat. The buzz helps when pitching fintech partners such as Chime and Klarna, which use VantageScore 4.0 for personal-loan decisions.

Still, the tactic irritates mortgage insiders. “Comparing your 2017 model to 1989 code is like bragging your Tesla outsells a Ford Model T,” quipped one senior credit-risk officer at a top-10 lender who spoke on condition of anonymity because he negotiates with both scoring firms.

Expect the comparison to persist. VantageScore’s 2024 investor pitch book allocates 18 of 45 slides to Classic FICO head-to-head charts, but only two address 10T, and both cite public-relations talking points rather than independent data.

What the 2026 Review Could Change

Inside the next test cycle and why both firms are already coding

FHFA’s 2022 approval came with a caveat: re-validation must occur by January 2026. Both FICO and VantageScore have begun building next-gen models. FICO’s working name is 10T NG; VantageScore’s is 5.0, according to people familiar with the roadmaps who asked not to be named because the brands are not yet public.

The 2026 test will add two new datasets: real-time cash-flow underwriting data from bank accounts and specialty “alternative data” such as rent, utility and telecom payments. FHFA wants to see whether either model can safely score an additional 4.8 million credit-invisible potential homebuyers without weakening loss reserves.

Early simulations by the Urban Institute show cash-flow data can lift predictive power another 7-9%, but fair-lending risk rises if algorithms treat irregular gig income differently across racial groups. Expect both firms to submit models that include bias-mitigation layers, a feature absent from today’s 10T and 4.0.

Whichever model wins stays exclusive until 2031, controlling a mortgage market that the Mortgage Bankers Association forecasts at $2.6 trillion annually.

Alternative data availability has exploded since 2022. Plaid, Yodlee and Finicity now connect to 200 million consumer bank accounts, offering 24 months of transaction history. Early tests show rent-payment tracking can verify on-time records for 9.7 million households that currently lack mortgage scores, according to the Urban Institute.

Yet consumer-protection groups warn that cash-flow data can encode income volatility in ways that disadvantage Black and Latino borrowers, who are twice as likely to work in gig sectors. FHFA has therefore proposed a two-step test: first, prove the model raises approval rates for protected classes; second, show no statistically significant increase in default rate gaps.

Both scoring firms have opened “responsible AI” divisions. FICO hired former CFPB chief data scientist Angela Lacerda in 2023; VantageScore partnered with the National Fair Housing Alliance. Each side must submit algorithmic-impact statements alongside their 2026 filings, a first for U.S. credit scoring.

Meanwhile, Capitol Hill is watching. A bipartisan bill introduced in May 2024 would require FHFA to approve at least two competing models if both pass safety-and-soundness tests. The provision, tucked into the larger Housing Supply Act, has already cleared committee and could become law before the 2026 review, reopening the market to VantageScore even if 10T NG retains a performance edge.

Bottom Line for Consumers Shopping Mortgages in 2024

Rate quotes, approval odds and the 620 cliff

If you plan to buy or refinance with a conforming loan in 2024, your lender will pull FICO Score 10T—period. The score you see on most free apps is Classic FICO 8, so expect a delta, usually ±10 points, says Keith Gumbinger, vice president of HSH.com.

More important is the 620 cliff. Below that mark, Fannie Mae’s automated underwriting system rejects all loans with less than 25% down. FHFA data show 10T pushes 4.3% of applicants under 620 who would have cleared 640 under Classic FICO, but lifts 5.1% of thin-file borrowers above the threshold when utility bills are counted.

The net effect: slightly tighter credit at the margin, but 190,000 additional approvals annually because 10T’s lower false-positive rate offsets the stricter scoring. For rate shoppers, the spread between 620-639 and 740-759 tiers averages 97 basis points on a 30-year fixed, translating to $210 per month on a $350,000 loan.

Bottom line: pay down revolving balances six months before application and let Experian Boost pull in on-time phone and utility payments. Those two moves alone can add 30-40 points under 10T’s trended logic.

Timing matters. Because 10T looks back 30 months, paying off a card today will not fully register for two billing cycles. Mortgage brokers routinely tell clients to start “credit hygiene” at least 45 days before pre-approval, up from 30 days under Classic FICO.

Also check for medical-collection errors. A 2023 CFPB audit found 59% of medical collections under $500 were wrongly reported; disputing them raised median 10T scores by 14 points, enough to move a borrower from 616 to 630, often the difference between denial and a 3% down payment.

Finally, do not obsess over the exact number. Once above 740, lenders offer the same par rate; the average 750 borrower receives the same quote as an 800 borrower, according to Optimal Blue pricing engine data. Focus on crossing the 620, 660 and 740 thresholds rather than chasing an 850.