We tried to rebuild S&P's "best-performing bank" ranking from public data. It didn't reproduce the real result — and the reason is more interesting than a simple mismatch.

Direct answer

S&P Global Market Intelligence named Central Bancompany Inc. its #1 U.S. public bank with more than $10 billion in assets for the 2025 ranking cycle, the only bank to beat the analysis median on all of what S&P calls its "seven-ranking metrics" for that tier — without naming which seven. [C01] S&P has separately, fully documented an 8-metric weighted framework in its community-bank-ranking material. [C02] Whether that is the same framework used for the public-bank tier, minus one metric, or a materially different one, is not resolved by any public source. [C03] Building the best-documented version anyway and applying it to OptimaYield's own $10B+ Call Report population, the replication does not reproduce the real result: Central Trust Bank, the bank charter beneath Central Bancompany, ranks 35th of 135 institutions, not first. [C04, C06] Narrowing the population to exclude obviously ineligible charter types barely moves that rank (30th of 120), so the gap is not primarily a population-eligibility problem. [C08] The most likely remaining explanations — ranking the bank charter instead of the holding company, and simplified metric definitions — are disclosed but not confirmed. [C09, C10]

Key findings

What's actually public about S&P's methodology — and what isn't

S&P Global Market Intelligence has disclosed, in its community-bank-ranking material, an exact 8-metric weighted scheme: pretax return on average assets (25%), net interest margin (10%), efficiency ratio (10%), 3-year average operating revenue change (10%), 8-quarter average deposit change (10%), nonperforming assets and 90-day-past-due loans as a share of total assets (10%), net charge-offs to average loans (10%), and adjusted tangible common equity to tangible assets (15%) — each standardized by standard deviation from the industry mean, weighted, summed, with caps and floors applied per metric to limit outlier influence. [C02]

That is a real, citable, exact methodology — for the community-bank category. The >$10B public-bank category's own 2025-cycle release describes Central Bancompany as beating the median on all "seven-ranking metrics," not eight, and does not say which seven. [C01] No public source resolves whether the public-bank tier drops one of the eight community-bank metrics, uses a different set entirely, or whether "seven" was even metric count in a strict sense. This is disclosed here as a real gap in what can be independently verified, rather than assumed away by treating the community-bank framework as if it were confirmed to be identical. [C03]

A ranking can be exactly reproducible in principle and still not reproducible in practice, if the tier-specific version of its own stated methodology isn't the one that's actually public.

Building the best-available replication anyway — and where it lands

Given that gap, this analysis uses the fully documented 8-metric framework as the closest available public proxy, computed from OptimaYield's own $10B+ detailed Call Report population using already-validated field-code logic (reusing deriveQuarterlyReserveSeries for net charge-offs rather than re-deriving roughly twenty RIBI category fields by hand). 135 of 159 institutions have all 8 metrics derivable as of June 30, 2026. [C04]

The resulting top of the ranking looks structurally different from what "best-performing bank" usually connotes: Beal Bank USA, Ameriprise Bank FSB, Optum Bank, BMW Bank of North America, American Express National Bank, Synchrony Bank, Atlantic Union Bank, Axos Bank, Barclays Bank Delaware, and Sallie Mae Bank fill the top 10 — a mix of monoline card, wealth-affiliated, auto-finance, and industrial-bank charters with unusually high margins or thin branch-driven cost structures, not diversified regional or community banks. [C05]

Central Trust Bank — the Call Report charter beneath Central Bancompany Inc., S&P's real #1 — ranks 35th of 135 in this replication, with a composite score of 0.176 against a population median of -0.054. [C06] It is not a weak performer: it beats the population median on 6 of the 8 metrics (pretax ROAA, NIM, efficiency ratio, 3-year revenue growth, NPA ratio, net charge-off ratio). It trails the median specifically on 8-quarter average deposit growth (1.25% vs. 1.32%) and adjusted TCE ratio (7.41% vs. 9.35%) — those two metrics, not overall weakness, are what keep it out of the top tier of this replication. [C07]

Testing one explanation directly — and ruling it out as sufficient

The most obvious candidate explanation is population eligibility: maybe several of the top-10 names above (monoline card banks, wealth-management-affiliated banks, industrial banks) aren't the kind of institution S&P's "public bank" ranking actually intends to include, and a narrower, more S&P-like population would push Central Trust Bank up. Excluding 15 likely foreign-parent-owned, trust-only, or custody-charter names (a disclosed, non-exhaustive list, not S&P's own eligibility criteria) narrows the population to 120. Central Trust Bank's rank moves from 35th of 135 to 30th of 120 — barely, and the top-10 list is nearly unchanged. [C08] Eligibility filtering, at least of this kind, is not sufficient on its own to explain the gap.

Two disclosed possibilities remain untested here: this replication ranks the bank charter (Central Trust Bank) rather than the publicly traded holding company (Central Bancompany Inc.) that S&P actually ranks — holding-company-level leverage, preferred stock, or non-bank subsidiary revenue could shift several of the 8 metrics in either direction, and this analysis does not have consolidated holding-company financials to test that directly. [C09] Separately, this replication's own metric definitions are disclosed simplifications — NIM on average total assets rather than average earning assets, no preferred-stock backout in the TCE ratio, an independently chosen ±3 standard-deviation outlier cap — not verified matches to S&P's exact, undisclosed parameters. [C10] Either factor, or both together, could plausibly account for the remaining gap; neither is confirmed.

Methodology

Population: every institution in OptimaYield's detailed ($10B+) Call Report layer with all 8 metrics of the documented community-bank framework derivable as of 2026-06-30 (135 of 159). Metrics reuse already-validated field-code branching (RCFD/RCON prefixes for consolidated vs. domestic-only filers) from validated-metric-contracts.ts's own pattern, and reuse deriveQuarterlyReserveSeries (from the CECL-adequacy work) for net charge-offs rather than re-deriving the underlying RIBI charge-off/recovery detail a second time. Standardization follows S&P's disclosed approach: each metric is centered on the population mean, scaled by standard deviation, capped/floored at ±3 SD (OptimaYield's own choice, not S&P's undisclosed exact parameter), weighted per C02's disclosed weights, and summed. Trailing metrics (3-year revenue growth, 8-quarter deposit growth) use year-end 2022-2025 and the trailing nine quarters through 2026-06-30 — the most recent full window OptimaYield's own data supports, not necessarily aligned to S&P's own 2025-cycle window. Live query: scripts/analysis-sp-best-performer-replication.ts.

Limitations and counter-evidence

• The exact public-bank-tier methodology is not public. [C03] This replication necessarily substitutes the best-documented adjacent framework (community banks) for the tier's own undisclosed "seven-ranking metrics."
• Entity-level mismatch is untested. [C09] Ranking the bank charter instead of the holding company is a real, disclosed scope difference this analysis cannot resolve without consolidated holding-company financials.
• Metric definitions are disclosed simplifications, not verified matches to S&P's exact ones. [C10]
• Population-eligibility filtering was tested and found insufficient on its own. [C08] This is reported as a negative result, not omitted because it didn't move the needle.
• This is a methodology-sensitivity test, not a claim that S&P's ranking is wrong. A ranking built on undisclosed exact parameters cannot be definitively invalidated by an independent replication that also has to guess at some of its own — the honest conclusion is that the replication is fragile to exactly the choices that aren't public, not that the real ranking is mistaken.
• Two tiers below $10B (community banks $3-10B, and under $3B) are out of scope. OptimaYield's detailed Call Report layer only covers $10B+ institutions; extending this test to the tiers where the full 8-metric methodology is actually documented is a natural next step, not attempted here.

Related on OptimaYield

• Is loan growth actually "broad-based"? — the same discipline of testing a widely cited industry characterization against a specific, falsifiable, disclosed definition rather than accepting it at face value.
• The bank-profile pages for the named institutions above (not yet linked pending publication approval).

Official sources

• S&P Global Market Intelligence, "S&P Global Market Intelligence Releases Annual Rankings of Best-Performing U.S. Community Banks, Public Banks and Credit Unions for 2025," press.spglobal.com, March 2026.
• 1st Source Corporation, "1st Source Again Ranked in Top 50 Community Banks by S&P Global," citing S&P Global Market Intelligence's disclosed 2024-cycle methodology.