Library/Fundamental Analysis and Valuation/Sector-Specific Equity Scoring/Insurance Fundamental Score

D18-F10-A02 / Complete engineering topic

Insurance Fundamental Score

Construct a Solvency-II-tagged insurer score that keeps capital requirements and underwriting economics separately inspectable.

Insurance Fundamental Score routes sector evidence into an auditable component and decision ledgerD18 / D18-F10

Construct a Solvency-II-tagged insurer score that keeps capital requirements and underwriting economics separately inspectable.

Insurance Fundamental Score route, calculation, and decision map

The decision this tutorial makes visible

Insurance liabilities, regulatory capital, reserve revisions, and contract accounting require a model different from banks and industrial firms.

The precise question is: How can solvency, underwriting, reserve development, own-fund quality, concentration, and profitability be combined for a non-life insurer?

A practitioner needs to know what the diagnostic does and does not justify. A builder needs a contract that can be reproduced from the same point-in-time inputs in Python, TypeScript, a visual, and a browser lab.

Intuition before notation

Solvency is the shock absorber, underwriting is the engine, and reserve development tests whether prior estimates are holding up.

The result depends on the declared algorithm scope, input clocks, units, equality and rounding policies, and unsupported-state treatment. Change one of those and the output represents a different decision even when its field name is unchanged.

Scope and nearby methods

This version covers a synthetic Solvency II non-life insurer. SCR and MCR coverage below 100% create an explicit review state and penalty; cross-regime comparisons require rerouting.

VariantDefinitionBest useMain limitation
Selected non-life Solvency IISolvency, underwriting and reserve panelEU-style non-life reviewNot portable to every regime
Life-insurer modelDuration, guarantees and asset-liability matchingLife and annuity writersNot represented here
Bank capital modelRisk-weighted assets and funding liquidityDeposit-taking bankWrong liability economics

What is sourced, selected, synthetic, and derived

RoleMaterial claimEvidenceBoundary
sourced factSolvency II distinguishes SCR, MCR, eligible own funds, and tiers.EIOPA Articles 101 and 297This composite is not part of Solvency II.
sourced factNAIC defines combined ratio as loss ratio plus expense ratio and uses reserve development in P/C regulatory screening.NAIC glossary and IRIS P/C ratio manualThose U.S. concepts do not make NAIC and Solvency II regimes interchangeable or endorse package bands.
sourced factIFRS 17 gives insurance contracts a specialized accounting model.IFRS FoundationAccounting results are not identical to prudential capital.
package choiceSeven bands and weights create the teaching score.Repository contractThey are neither regulatory nor actuarial standards.
derived outputThe canonical score is a deterministic review aid.Calculation ledgerIt is not a solvency opinion or recommendation.

The authoritative sources support only the exact facts named in the claim ledger. They do not certify the synthetic numbers in this tutorial. The repository fixture is deliberately invented for auditability, and the displayed output is author-derived under the selected implementation choice.

Formula, symbols, and numerical policy

Plain text
I = clamp(Σ wᵢ sᵢ − 20 × number of SCR/MCR coverage breaches, 0, 100).
SymbolMeaningUnitPolicy
SCRcSCR coveragemultipleKeep regime and group/solo scope
CRcombined ratiodecimal ratioLower is better within package band
ARDadverse reserve developmentdecimal ratioAlign reserve basis and horizon
  • Use IEEE-754 binary64 arithmetic without intermediate rounding.
  • Treat percentages as decimal fractions and label percentage-point changes explicitly.
  • Clamp only declared component transforms to [0,100]; never clamp raw regulatory or accounting inputs silently.
  • Round only for presentation after the full component, penalty, and coverage ledger is stored.

Read the formula in the same order as the algorithm. Validate identity, ordering, units, and supported state first. Apply the selected equality and window rules second. Calculate with unrounded numeric values. Round only at the declared presentation boundary, and preserve null as a diagnostic rather than coercing it to zero.

Build the algorithm

  1. Confirm non-life business mix, Solvency II scope, group perimeter, and cutoff.
  2. Normalize the seven raw measures without crossing accounting and regulatory scopes.
  3. Transform each measure through the declared band.
  4. Apply weights and the explicit SCR/MCR breach penalty.
  5. Publish score, state, breaches, components, and limitations.

Production-minded operational checklist

  1. Resolve sector and framework before calculating.
  2. Freeze scoring and knowledge-cutoff timestamps.
  3. Reconcile each input to its filing, regulator, provider, and reporting perimeter.
  4. Inspect component scores, penalties, and coverage before the headline.
  5. Retain abstention and unsupported-scope reasons for the integrated router.

The checklist is intentionally strict: an explicit rejection is safer than a plausible output built from stale, malformed, or unsupported state.

Worked synthetic example

The canonical fixture is synthetic teaching data, not an observed control event, customer order, or broker execution. Its primary author-derived output, fundamental_score, is 73.614358974359. The complete input and output are in datasets/canonical-input.json and datasets/expected-output.json.

The canonical synthetic insurer clears SCR and MCR coverage and reports a combined ratio below 100%. The result is the weighted seven-component ledger with no capital penalty.

Counterfactual checkpoint

Stress the combined ratio. Hold all fields fixed and increase the combined ratio through 100% toward 108%. The output changes because The underwriting component weakens while solvency remains unchanged, revealing two different causal channels.

The structured result retains state and diagnostics in addition to the primary number. That makes the calculation independently reviewable and prevents a partial, null, rejected, or venue-bounded outcome from being mistaken for an unqualified value.

Boundary and counterexample workbook

The playground computes every scenario at 61 deterministic parameter states. The table uses the declared focus step and states whether that focus reproduces the canonical fixture. The full state ledger and compressed transition segments are in datasets/scenario-results.json.

ScenarioReview focusPurposeStatePrimary outputDiagnosticDecision segments
Canonical driver sweepStep 30 · canonical fixtureMove one primary causal driver through the canonical midpoint.mixed-review-bandfundamental score 73.6mixed-review-band · solvency-ii-tagged-nonlife-package-bands2
Adverse operating stressStep 30 · comparison focusApply a controlled operating or earnings stress while retaining the framework.mixed-review-bandfundamental score 69.4mixed-review-band · solvency-ii-tagged-nonlife-package-bands2
Balance-sheet resilienceStep 30 · comparison focusVary a funding, leverage, capital, or liquidity channel.mixed-review-bandfundamental score 69.3mixed-review-band · solvency-ii-tagged-nonlife-package-bands3
Boundary crossingStep 30 · comparison focusCross a declared review boundary and inspect equality behavior.mixed-review-bandfundamental score 69.9mixed-review-band · solvency-ii-tagged-nonlife-package-bands2
Evidence-quality stressStep 30 · comparison focusChange evidence usability or freshness without hiding the diagnostic.mixed-review-bandfundamental score 71.2mixed-review-band · solvency-ii-tagged-nonlife-package-bands2
Concentration or mixStep 30 · comparison focusVary concentration, mix, or component balance.mixed-review-bandfundamental score 71.5mixed-review-band · solvency-ii-tagged-nonlife-package-bands2
Recovery pathStep 30 · comparison focusTrace a recovery from an adverse state toward a resilient state.mixed-review-bandfundamental score 71.3mixed-review-band · solvency-ii-tagged-nonlife-package-bands2

These rows are not backtest observations. They are controlled counterexamples that expose how one driver changes the state, output, or reason code while the rest of the contract stays fixed.

Visualize the boundary

Insurance Fundamental Score annotated teaching map

Open this SVG at full size, or use the guided playground to compare the seven topic-specific canonical, boundary, policy, and failure scenarios.

The Mermaid flow answers where the selected calculation sits in the processing sequence. The SVG keeps the formula, output, decision boundary, and invariant visible together. The lab lets the reader step through the same structured states without changing the underlying definition.

Implementation walkthrough

The Python and TypeScript references begin with the same validation contract, reject malformed and unsupported state before calculation, preserve declared ordering and rounding policies, and return structured diagnostics rather than one context-free number.

The main implementation branches are:

  • SCR or MCR coverage is below 1.0 — Emit capital-requirement-review and penalize, because Capital floors should not disappear in an average.
  • Life, health, composite, or non-Solvency-II scope dominates — Reroute or abstain, because The selected transform is non-life and regime-specific.
  • Definitions and perimeter reconcile — Publish score with capital and underwriting ledgers, because Reviewers need both views.

Neither reference silently fetches data, mutates caller-owned inputs outside the declared engine behavior, guesses hidden state, or substitutes a provider default. Shared JSON fixtures make value, null, state, and reason-code drift visible across languages.

Testing and validation

Definition tests compare every canonical field, reject malformed state, and exercise the material boundary. Family validation recomputes every playground state from the reference function. Independent arithmetic is recorded beside the fixture rather than inferred only from implementation output.

The audit must preserve these invariants:

  • framework and policy version
  • point-in-time normalized input ledger
  • component transform thresholds and scores
  • weight vector and penalties
  • coverage, state, reason, and final output

Passing the suite proves the selected package contract and Python/TypeScript parity; it does not prove empirical usefulness for a live issuer universe.

Failure modes and misuse

  • The 0-100 teaching bands and default weights are package choices, not regulations, ratings, or market standards.
  • Definition fidelity and code parity do not establish predictive validity, calibration, causality, fair value, or investment performance.
  • Accounting frameworks, prudential regimes, business mixes, and reporting perimeters can make apparently identical ratios incomparable.
  • A complete component ledger can still omit material qualitative, governance, legal, catastrophe, operational, or market risks.

Debugging order

When a result looks surprising, inspect the state in this order:

  1. Confirm framework and reporting perimeter
  2. Check units and knowledge dates
  3. Recalculate raw intermediates
  4. Inspect bounded component transforms
  5. Reconcile weights, penalties, coverage, and state

Evidence and historical boundary

Historical decision: not useful. A named Insurance Fundamental Score case would introduce issuer identity, filing and regulator scope, restatement, data-license, and scoring-policy questions without teaching the deterministic mechanism better than a controlled synthetic record.

The primary sources are Solvency II Article 101, Solvency II Article 297, IFRS 17, NAIC insurance glossary, NAIC IRIS P/C ratios, SEC non-GAAP C&DIs. They support the source roles listed in the research ledger, not a redistributable historical observation, a private participant decision, production conformance certification, execution-quality result, profitability claim, or prediction claim.

Summary and next topic

You can now compute and audit Insurance Fundamental Score before passing its versioned output to the integrated scoring router. The learning flow is: Bank Fundamental Score → Insurance Fundamental Score → REIT Fundamental Score. Carry the result forward only with its scope, clock, state, and evidence label.

Verified model contract — Level 1

  • Selected calculation: I = clamp(Σ wᵢ sᵢ − 20 × number of SCR/MCR coverage breaches, 0, 100).
  • Reproducibility boundary: the canonical fixture is synthetic; the stored input, intermediate ledger, output, and cross-language implementations define the executable example.
  • Interpretation ceiling: Definition fidelity and code parity do not establish predictive validity, calibration, causality, fair value, or investment performance.
  • Insurance evidence boundary: the NAIC combined ratio adds the loss and expense ratios, while IRIS reserve-development ratios are contextual regulatory screens. Neither supplies this package's transform bands or makes U.S. P/C measures interchangeable with Solvency II capital measures.

Level 2 learning layer

How to choose this model or a nearby method

MethodBest useAssumption that must holdMain limitation
Selected non-life Solvency IIEU-style non-life reviewComparable group and regulatory scopeNot portable to every regime
Life-insurer modelLife and annuity writersLiability options are modeledNot represented here
Bank capital modelDeposit-taking bankBank balance-sheet mechanicsWrong liability economics

The selected package is appropriate only when its framework tag, reporting perimeter, unit policy, and point-in-time evidence contract are all satisfied. If a stop condition in the topic decision matrix fires, use the named review, reroute, or abstention state instead of forcing a score.

Compact glossary

SymbolMeaningUnitPolicy
SCRcSCR coveragemultipleKeep regime and group/solo scope
CRcombined ratiodecimal ratioLower is better within package band
ARDadverse reserve developmentdecimal ratioAlign reserve basis and horizon

Visual asset map

AssetLearning jobStatus
Article heroSee route → normalize → calculate → decide at a glanceReady
System mapConnect formula, boundary, invariant, counterfactual, and outputReady
Model anatomyFollow the four-stage calculation pipelineReady
Component ledgerReconstruct the headline from named dimensionsReady
Decision boundarySeparate a computable result from a permissible interpretationReady
Evidence clockPrevent point-in-time leakageReady
Mermaid flowInspect the text-native algorithm flowReady
Guided playgroundPredict, reveal, sweep 427 calculations, and explain state changesReady

Related concepts and continuation

Deep visual atlas

Sector model anatomy

Canonical component ledger

Decision boundary and interpretation ceiling

Point-in-time evidence clock

The teaching sequence is route → normalize → calculate → decide. Use the component ledger to reconstruct the headline, the boundary map to stop unsupported interpretation, and the evidence clock to block hindsight.

Guided playground protocol

  1. Orient: identify the framework, active driver, and invariant.
  2. Predict: commit to decrease, stay flat, or increase before revealing the adjacent numeric result.
  3. Experiment: sweep all 61 computed states in each of seven scenarios and inspect discontinuities.
  4. Explain: reconcile underwriting component beside SCR/MCR breach checks to the headline and apply this boundary: Life-heavy, non-life-incompatible, or cross-regime records must reroute.

Insurance Fundamental Score calculation flow

This flow identifies the selected calculation stages and the structured output.

Rendering system map…

Takeaway: Insurer capital strength and underwriting quality answer different questions.

ReferencesPrimary sources and evidence notes

Expand the source trail, evidence role, and limitations behind the engineering choices.

S1 — Calculation of the Solvency Capital Requirement

  • Organization or authors: European Parliament and Council via EIOPA Single Rulebook
  • Source type: Official prudential law
  • Publication or effective date: Directive 2009/138/EC; accessed 2026-08-08
  • Version: Article 101
  • URL or DOI: https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-2188_en
  • Accessed: 2026-08-09
  • Jurisdiction: European Union Solvency II
  • Supports: SCR covers specified risks and is calibrated to a 99.5% one-year confidence level under the directive.
  • Limitations: Does not make a Solvency II ratio comparable with another regime or define this composite.

S2 — Capital management disclosures

  • Organization or authors: European Commission via EIOPA Single Rulebook
  • Source type: Official delegated regulation
  • Publication or effective date: Accessed 2026-08-08
  • Version: Article 297
  • URL or DOI: https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-5960_en
  • Accessed: 2026-08-09
  • Jurisdiction: European Union Solvency II
  • Supports: Solvency reporting distinguishes eligible own funds, tiers, SCR, MCR, and material differences from financial-statement equity.
  • Limitations: Does not validate underwriting or reserve-quality thresholds in this package.

S3 — IFRS 17 Insurance Contracts

  • Organization or authors: International Accounting Standards Board
  • Source type: Official accounting standard overview
  • Publication or effective date: Effective 2023; accessed 2026-08-08
  • Version: Current standard page
  • URL or DOI: https://www.ifrs.org/issued-standards/list-of-standards/ifrs-17-insurance-contracts/
  • Accessed: 2026-08-09
  • Jurisdiction: IFRS
  • Supports: Insurance-contract measurement and presentation differ from ordinary industrial-company accounting.
  • Limitations: The public overview does not define a universal insurer equity score.

S4 — Glossary of Insurance Terms — Combined Ratio

  • Organization or authors: National Association of Insurance Commissioners
  • Source type: Official regulator association glossary
  • Publication or effective date: Accessed 2026-08-09
  • Version: Current web glossary
  • URL or DOI: https://content.naic.org/glossary-insurance-terms
  • Accessed: 2026-08-09
  • Jurisdiction: United States insurance regulation
  • Supports: The combined ratio is calculated by adding the loss and expense ratios and indicates underwriting profitability.
  • Limitations: It is not a Solvency II capital ratio and the package's 0-100 underwriting transform is not an NAIC threshold.

S5 — Insurance Regulatory Information System — Property/Casualty Ratios

  • Organization or authors: National Association of Insurance Commissioners
  • Source type: Official regulator analysis manual
  • Publication or effective date: 2024 edition; accessed 2026-08-09
  • Version: 2024 UIR-ZB-24
  • URL or DOI: https://content.naic.org/sites/default/files/uir-zb-24.pdf
  • Accessed: 2026-08-09
  • Jurisdiction: United States property/casualty regulatory screening
  • Supports: IRIS uses one- and two-year reserve-development ratios as regulatory screening inputs and directs analysts to interpret unusual results in context.
  • Limitations: The package uses a different normalized adverse-development input and must not inherit IRIS ranges or imply a U.S. regulatory conclusion.

S6 — Non-GAAP Financial Measures Compliance and Disclosure Interpretations

  • Organization or authors: U.S. Securities and Exchange Commission
  • Source type: Official staff guidance
  • Publication or effective date: Updated 2022; accessed 2026-08-08
  • Version: Current web version
  • URL or DOI: https://www.sec.gov/rules-regulations/staff-guidance/corporation-finance-interpretations/non-gaap-financial-measures
  • Accessed: 2026-08-09
  • Jurisdiction: United States SEC reporting
  • Supports: Adjusted measures can be misleading when definitions, recurring adjustments, prominence, or reconciliation are inadequate.
  • Limitations: Does not standardize analyst AFFO, utility FFO, or normalized commodity earnings.

Evidence boundary

Sources define sector measures, disclosure regimes, and methodological cautions. They do not endorse the repository's synthetic fixtures, weights, score bands, or investment use.

sector-scoring.ts
/** Deterministic TypeScript reference for D18-F10 sector-specific scoring. */

export type RecordValue = Record<string, any>;

const numberValue = (data: RecordValue, name: string): number => {
  const value = data[name];
  if (typeof value !== "number" || !Number.isFinite(value)) throw new TypeError(`${name} must be a finite number`);
  return value;
};
const nonnegative = (data: RecordValue, name: string): number => {
  const value = numberValue(data, name); if (value < 0) throw new RangeError(`${name} must be nonnegative`); return value;
};
const positive = (data: RecordValue, name: string): number => {
  const value = numberValue(data, name); if (value <= 0) throw new RangeError(`${name} must be positive`); return value;
};
const textValue = (data: RecordValue, name: string): string => {
  const value = data[name]; if (typeof value !== "string" || !value.trim()) throw new TypeError(`${name} must be a nonempty string`); return value.trim();
};
const dateValue = (value: unknown, name: string): Date => {
  if (typeof value !== "string" || !/^\d{4}-\d{2}-\d{2}$/.test(value)) throw new TypeError(`${name} must be YYYY-MM-DD`);
  const result = new Date(`${value}T00:00:00Z`); if (!Number.isFinite(result.getTime()) || result.toISOString().slice(0, 10) !== value) throw new RangeError(`${name} must be YYYY-MM-DD`); return result;
};
const requireFramework = (data: RecordValue, expected: string): void => { if (textValue(data, "framework") !== expected) throw new RangeError(`framework must be ${expected}`); };
const clamp = (value: number, low = 0, high = 100): number => Math.min(high, Math.max(low, value));
const higher = (value: number, weak: number, strong: number): number => { if (strong <= weak) throw new RangeError("strong threshold must exceed weak threshold"); return clamp(100 * (value - weak) / (strong - weak)); };
const lower = (value: number, strong: number, weak: number): number => { if (weak <= strong) throw new RangeError("weak threshold must exceed strong threshold"); return clamp(100 * (weak - value) / (weak - strong)); };
const centered = (value: number, center: number, fullDistance: number, zeroDistance: number): number => {
  if (zeroDistance <= fullDistance) throw new RangeError("zero-score distance must exceed full-score distance");
  return clamp(100 * (zeroDistance - Math.abs(value - center)) / (zeroDistance - fullDistance));
};
const weightsFor = (data: RecordValue, names: string[]): Record<string, number> => {
  const raw = data.weights;
  if (!raw || typeof raw !== "object" || Array.isArray(raw) || Object.keys(raw).sort().join("|") !== [...names].sort().join("|")) throw new RangeError("weights must contain exactly the declared component names");
  const result = Object.fromEntries(names.map(name => [name, nonnegative(raw, name)]));
  const total = Object.values(result).reduce((sum, value) => sum + value, 0);
  if (Math.abs(total - 1) > 1e-9) throw new RangeError("weights must sum to 1");
  return result;
};
const weightedScore = (components: Record<string, number>, weights: Record<string, number>): number => Object.entries(components).reduce((sum, [name, value]) => sum + value * weights[name], 0);
const scoreState = (score: number): string => score >= 75 ? "strong-review-band" : score >= 50 ? "mixed-review-band" : "weak-review-band";

function bankScore(data: RecordValue): RecordValue {
  requireFramework(data, "basel-iii-teaching-v1");
  const names = ["capital", "leverage", "short_liquidity", "stable_funding", "asset_quality", "coverage", "margin", "efficiency"];
  const weights = weightsFor(data, names);
  const values = {cet1_ratio: nonnegative(data,"cet1_ratio"), leverage_ratio: nonnegative(data,"leverage_ratio"), lcr: nonnegative(data,"lcr"), nsfr: nonnegative(data,"nsfr"), npl_ratio: nonnegative(data,"npl_ratio"), provision_coverage_ratio: nonnegative(data,"provision_coverage_ratio"), net_interest_margin: numberValue(data,"net_interest_margin"), cost_income_ratio: nonnegative(data,"cost_income_ratio")};
  const components = {capital:higher(values.cet1_ratio,.07,.14), leverage:higher(values.leverage_ratio,.03,.06), short_liquidity:higher(values.lcr,1,1.4), stable_funding:higher(values.nsfr,1,1.3), asset_quality:lower(values.npl_ratio,.02,.08), coverage:higher(values.provision_coverage_ratio,.6,1.2), margin:higher(values.net_interest_margin,.01,.04), efficiency:lower(values.cost_income_ratio,.4,.7)};
  const minimums = data.minimums; if (!minimums || typeof minimums !== "object" || Array.isArray(minimums)) throw new TypeError("minimums must be an object");
  const checks: Record<string,number> = {cet1_ratio:nonnegative(minimums,"cet1_ratio"), leverage_ratio:nonnegative(minimums,"leverage_ratio"), lcr:nonnegative(minimums,"lcr"), nsfr:nonnegative(minimums,"nsfr")};
  const breaches = Object.entries(checks).filter(([name, minimum]) => values[name as keyof typeof values] < minimum).map(([name])=>name);
  const base = weightedScore(components, weights), penalty = 12.5 * breaches.length, score = clamp(base - penalty);
  return {state:breaches.length?"prudential-floor-review":scoreState(score),method:"bank-sector-score-v1",component_scores:components,weights,base_score:base,floor_breaches:breaches,penalty,fundamental_score:score,coverage_ratio:1,reason:"package-bands-with-declared-prudential-minimums"};
}

function insuranceScore(data: RecordValue): RecordValue {
  requireFramework(data,"solvency-ii-nonlife-teaching-v1");
  const names=["solvency","minimum_capital","underwriting","reserve_quality","own_fund_quality","concentration","profitability"], weights=weightsFor(data,names);
  const values={scr_coverage_ratio:nonnegative(data,"scr_coverage_ratio"),mcr_coverage_ratio:nonnegative(data,"mcr_coverage_ratio"),combined_ratio:nonnegative(data,"combined_ratio"),adverse_reserve_development_ratio:numberValue(data,"adverse_reserve_development_ratio"),tier1_own_funds_share:nonnegative(data,"tier1_own_funds_share"),investment_concentration_ratio:nonnegative(data,"investment_concentration_ratio"),return_on_equity:numberValue(data,"return_on_equity")};
  const components={solvency:higher(values.scr_coverage_ratio,1,2),minimum_capital:higher(values.mcr_coverage_ratio,1,3),underwriting:lower(values.combined_ratio,.88,1.08),reserve_quality:lower(values.adverse_reserve_development_ratio,-.02,.08),own_fund_quality:higher(values.tier1_own_funds_share,.5,.9),concentration:lower(values.investment_concentration_ratio,.1,.4),profitability:higher(values.return_on_equity,.02,.15)};
  const breaches=["scr_coverage_ratio","mcr_coverage_ratio"].filter(name=>values[name as keyof typeof values]<1),base=weightedScore(components,weights),penalty=20*breaches.length,score=clamp(base-penalty);
  return {state:breaches.length?"capital-requirement-review":scoreState(score),method:"insurance-sector-score-v1",component_scores:components,weights,base_score:base,floor_breaches:breaches,penalty,fundamental_score:score,coverage_ratio:1,reason:"solvency-ii-tagged-nonlife-package-bands"};
}

function reitScore(data: RecordValue): RecordValue {
  requireFramework(data,"nareit-equity-reit-teaching-v1");
  const names=["distribution","leverage","coverage","occupancy","same_store_growth","liquidity"],weights=weightsFor(data,names);
  const ffo=numberValue(data,"nareit_ffo"),capex=nonnegative(data,"recurring_capex"),rent=numberValue(data,"straight_line_rent_adjustment"),dividends=positive(data,"common_dividends"),ebitda=positive(data,"ebitda_re"),interest=positive(data,"interest_expense"),nearDebt=positive(data,"near_term_debt_maturities");
  const affo=ffo-capex-rent,distribution=affo/dividends,leverage=nonnegative(data,"net_debt")/ebitda,interestCoverage=ebitda/interest,liquidity=nonnegative(data,"available_liquidity")/nearDebt;
  const components={distribution:higher(distribution,.8,1.4),leverage:lower(leverage,3.5,8),coverage:higher(interestCoverage,1.5,5),occupancy:higher(nonnegative(data,"occupancy_ratio"),.8,.97),same_store_growth:higher(numberValue(data,"same_store_noi_growth"),-.05,.08),liquidity:higher(liquidity,.75,2)};
  const score=weightedScore(components,weights);
  return {state:affo<=0?"affo-proxy-deficit-review":scoreState(score),method:"reit-sector-score-v1",nareit_ffo:ffo,affo_proxy:affo,distribution_coverage:distribution,net_debt_to_ebitda_re:leverage,interest_coverage:interestCoverage,liquidity_coverage:liquidity,component_scores:components,weights,fundamental_score:score,coverage_ratio:1,reason:"nareit-ffo-plus-explicit-package-affo-proxy"};
}

function utilityScore(data: RecordValue): RecordValue {
  requireFramework(data,"ferc-regulated-electric-teaching-v1");
  const names=["earned_return","cash_debt","capital_structure","interest_coverage","capex_funding","regulatory_lag","rate_base_growth"],weights=weightsFor(data,names);
  const allowed=positive(data,"allowed_roe"),earned=numberValue(data,"earned_roe"),debt=positive(data,"total_debt"),ffoDebt=numberValue(data,"funds_from_operations")/debt,interestCoverage=numberValue(data,"ebit")/positive(data,"interest_expense"),capexFunding=numberValue(data,"cash_from_operations")/positive(data,"capital_expenditure");
  const components={earned_return:centered(earned-allowed,0,.005,.04),cash_debt:higher(ffoDebt,.08,.22),capital_structure:lower(nonnegative(data,"debt_to_capital"),.4,.65),interest_coverage:higher(interestCoverage,1.5,5),capex_funding:higher(capexFunding,.4,1),regulatory_lag:lower(nonnegative(data,"regulatory_lag_months"),3,18),rate_base_growth:centered(numberValue(data,"rate_base_growth"),.05,.01,.08)};
  const score=weightedScore(components,weights);
  return {state:scoreState(score),method:"utility-sector-score-v1",earned_allowed_roe_gap:earned-allowed,ffo_to_debt:ffoDebt,interest_coverage:interestCoverage,capex_funding_ratio:capexFunding,component_scores:components,weights,fundamental_score:score,coverage_ratio:1,reason:"ferc-tagged-package-bands"};
}

function earlyStageScore(data: RecordValue): RecordValue {
  requireFramework(data,"early-stage-liquidity-teaching-v1");
  const names=["runway","burn_trend","obligation_cover","revenue_cover"],weights=weightsFor(data,names),cashFlows=data.monthly_net_cash_flows,revenues=data.monthly_revenue;
  if(!Array.isArray(cashFlows)||cashFlows.length<6)throw new RangeError("monthly_net_cash_flows must contain at least 6 observations");
  if(!Array.isArray(revenues)||revenues.length!==cashFlows.length)throw new RangeError("monthly_revenue must match cash-flow history");
  const flows=cashFlows.map((v:any)=>numberValue({v},"v")),revs=revenues.map((v:any)=>nonnegative({v},"v"));
  const unrestricted=nonnegative(data,"unrestricted_cash"),investments=nonnegative(data,"liquid_investments"),facility=nonnegative(data,"unconditionally_committed_facility"),obligations=nonnegative(data,"near_term_obligations"),minimumCash=nonnegative(data,"minimum_operating_cash");
  const available=unrestricted+investments+facility-obligations-minimumCash,latest=flows.slice(-6),burn=Math.max(0,-latest.reduce((a:number,b:number)=>a+b,0)/latest.length),prior=Math.max(0,-latest.slice(0,3).reduce((a:number,b:number)=>a+b,0)/3),current=Math.max(0,-latest.slice(3).reduce((a:number,b:number)=>a+b,0)/3),burnRatio=prior>0?current/prior:(current===0?0:2),gross=unrestricted+investments+facility,obligationCover=obligations>0?gross/obligations:10,latestRevenue=revs.slice(-3).reduce((a:number,b:number)=>a+b,0)/3,revenueCover=current>0?latestRevenue/current:10,runway=burn>0?available/burn:null;
  const components={runway:runway===null?100:higher(runway,6,24),burn_trend:lower(burnRatio,.75,1.25),obligation_cover:higher(obligationCover,.75,2),revenue_cover:higher(revenueCover,0,1)},score=weightedScore(components,weights);
  const state=available<0?"negative-available-liquidity":runway===null?"nonburning-observation":runway<12?"funding-window-under-12-months":scoreState(score);
  return {state,method:"early-stage-runway-score-v1",available_liquidity:available,monthly_burn:burn,prior_three_month_burn:prior,latest_three_month_burn:current,burn_ratio:burnRatio,runway_months:runway,obligation_coverage:obligationCover,revenue_to_burn:revenueCover,component_scores:components,weights,runway_score:score,coverage_ratio:1,reason:"historical-burn-sensitivity-not-management-forecast"};
}

const medianValue=(values:number[]):number=>{const sorted=[...values].sort((a,b)=>a-b),middle=Math.floor(sorted.length/2);return sorted.length%2?sorted[middle]:(sorted[middle-1]+sorted[middle])/2;};
function cyclicalScore(data:RecordValue):RecordValue{
  requireFramework(data,"cyclical-midcycle-teaching-v1");
  const names=["normalized_margin","normalized_leverage","cash_consistency","cycle_balance"],weights=weightsFor(data,names),arrays:Record<string,number[]>={};
  for(const name of ["realized_price_history","unit_cost_history","volume_history","free_cash_flow_history"]){const raw=data[name];if(!Array.isArray(raw)||raw.length<7)throw new RangeError(`${name} must contain at least 7 observations`);arrays[name]=raw.map((v:any)=>numberValue({v},"v"));}
  if(new Set(Object.values(arrays).map(v=>v.length)).size!==1)throw new RangeError("cycle histories must have equal length");
  const prices=arrays.realized_price_history,costs=arrays.unit_cost_history,volumes=arrays.volume_history;if([...prices,...costs,...volumes].some(v=>v<=0))throw new RangeError("price, cost, and volume histories must be positive");
  const normalizedPrice=medianValue(prices),normalizedCost=medianValue(costs),normalizedVolume=medianValue(volumes),revenue=normalizedPrice*normalizedVolume,ebitda=(normalizedPrice-normalizedCost)*normalizedVolume-nonnegative(data,"fixed_costs"),margin=ebitda/revenue,netDebt=nonnegative(data,"net_debt"),leverage=ebitda>0?netDebt/ebitda:null,positiveFcf=arrays.free_cash_flow_history.filter(v=>v>0).length/arrays.free_cash_flow_history.length,currentPrice=positive(data,"current_realized_price"),percentile=prices.filter(v=>v<=currentPrice).length/prices.length;
  const components={normalized_margin:higher(margin,0,.3),normalized_leverage:leverage===null?0:lower(leverage,.5,4),cash_consistency:higher(positiveFcf,.3,.9),cycle_balance:centered(percentile,.5,.1,.5)},score=weightedScore(components,weights);
  return {state:ebitda<=0?"nonpositive-midcycle-ebitda-review":scoreState(score),method:"cyclical-midcycle-normalization-v1",normalized_price:normalizedPrice,normalized_unit_cost:normalizedCost,normalized_volume:normalizedVolume,normalized_revenue:revenue,normalized_ebitda:ebitda,normalized_margin:margin,normalized_net_leverage:leverage,positive_fcf_ratio:positiveFcf,cycle_percentile:percentile,component_scores:components,weights,normalized_score:score,coverage_ratio:1,reason:"point-in-time-median-cycle-window"};
}

function holdingCompanyScore(data:RecordValue):RecordValue{
  requireFramework(data,"holding-company-lookthrough-teaching-v1");
  const names=["nav_buffer","lookthrough_leverage","parent_coverage","diversification","freshness","listed_coverage"],weights=weightsFor(data,names),holdings=data.holdings;
  if(!Array.isArray(holdings)||holdings.length<2)throw new RangeError("holdings must contain at least two records");
  const ids=new Set<string>(),rows=holdings.map((raw:any)=>{if(!raw||typeof raw!=="object"||Array.isArray(raw))throw new TypeError("each holding must be an object");const id=textValue(raw,"id");if(ids.has(id))throw new RangeError("holding IDs must be unique");ids.add(id);const ownership=numberValue(raw,"ownership_ratio");if(!(ownership>0&&ownership<=1))throw new RangeError("ownership_ratio must be in (0, 1]");if(typeof raw.listed!=="boolean"||typeof raw.stale!=="boolean")throw new TypeError("listed and stale must be booleans");return{id,ownership_ratio:ownership,attributable_equity_value:ownership*nonnegative(raw,"equity_value"),attributable_debt:ownership*nonnegative(raw,"debt"),attributable_dividends:ownership*nonnegative(raw,"dividends_to_parent"),listed:raw.listed,stale:raw.stale};});
  const parentCash=nonnegative(data,"parent_cash"),parentDebt=nonnegative(data,"parent_debt"),other=nonnegative(data,"other_parent_liabilities"),parentInterest=positive(data,"parent_interest_expense"),stake=rows.reduce((s,r)=>s+r.attributable_equity_value,0),gav=stake+parentCash;if(gav<=0)throw new RangeError("gross asset value must be positive");
  const nav=gav-parentDebt-other,attributableDebt=rows.reduce((s,r)=>s+r.attributable_debt,0),enterprise=stake+attributableDebt+parentCash,leverage=enterprise>0?(parentDebt+attributableDebt)/enterprise:1,parentCoverage=rows.reduce((s,r)=>s+r.attributable_dividends,0)/parentInterest,shares=stake>0?rows.map(r=>r.attributable_equity_value/stake):[1],concentration=shares.reduce((s,v)=>s+v*v,0),freshness=stake>0?rows.filter(r=>!r.stale).reduce((s,r)=>s+r.attributable_equity_value,0)/stake:0,listed=stake>0?rows.filter(r=>r.listed).reduce((s,r)=>s+r.attributable_equity_value,0)/stake:0,navRatio=nav/gav;
  const components={nav_buffer:higher(navRatio,.2,.8),lookthrough_leverage:lower(leverage,.2,.65),parent_coverage:higher(parentCoverage,1,4),diversification:lower(concentration,.25,.7),freshness:higher(freshness,.6,1),listed_coverage:higher(listed,.3,1)},score=weightedScore(components,weights),marketCap=data.parent_market_cap===undefined?null:nonnegative(data,"parent_market_cap"),discount=marketCap===null||nav<=0?null:1-marketCap/nav;
  return {state:nav<=0?"nonpositive-nav-review":scoreState(score),method:"holding-company-lookthrough-score-v1",holding_ledger:rows,gross_asset_value:gav,net_asset_value:nav,attributable_subsidiary_debt:attributableDebt,lookthrough_leverage:leverage,parent_interest_coverage:parentCoverage,concentration_hhi:concentration,fresh_value_coverage:freshness,listed_value_coverage:listed,discount_to_nav:discount,component_scores:components,weights,look_through_score:score,coverage_ratio:1,reason:"attributable-stakes-with-parent-bridge"};
}

const sigmoid=(value:number):number=>value>=0?1/(1+Math.exp(-value)):Math.exp(value)/(1+Math.exp(value));
const projectSimplex=(values:number[]):number[]=>{const ordered=[...values].sort((a,b)=>b-a);let cumulative=0,rho=0;ordered.forEach((value,index)=>{cumulative+=value;if(value-(cumulative-1)/(index+1)>0)rho=index+1;});const theta=(ordered.slice(0,rho).reduce((a,b)=>a+b,0)-1)/rho,projected=values.map(v=>Math.max(v-theta,0)),total=projected.reduce((a,b)=>a+b,0);return projected.map(v=>v/total);};
const brier=(rows:number[][],labels:number[],weights:number[],intercept:number,slope:number):number=>rows.reduce((sum,row,index)=>{const score=row.reduce((s,v,j)=>s+weights[j]*v,0)/100,p=sigmoid(intercept+slope*(score-.5));return sum+(p-labels[index])**2;},0)/rows.length;
function calibrateWeights(data:RecordValue):RecordValue{
  requireFramework(data,"sector-weight-calibration-teaching-v1");
  const names=data.component_names,rowsRaw=data.feature_rows,labelsRaw=data.labels,datesRaw=data.observation_dates;
  if(!Array.isArray(names)||names.length<2||new Set(names).size!==names.length||!names.every((n:any)=>typeof n==="string"&&n))throw new RangeError("component_names must be unique nonempty strings");
  if(!Array.isArray(rowsRaw)||rowsRaw.length<12)throw new RangeError("feature_rows must contain at least 12 observations");if(!Array.isArray(labelsRaw)||labelsRaw.length!==rowsRaw.length)throw new RangeError("labels must match feature_rows");if(!Array.isArray(datesRaw)||datesRaw.length!==rowsRaw.length)throw new RangeError("observation_dates must match feature_rows");
  const rows=rowsRaw.map((row:any)=>{if(!Array.isArray(row)||row.length!==names.length)throw new RangeError("each feature row must match component_names");const values=row.map((v:any)=>numberValue({v},"v"));if(values.some((v:number)=>v<0||v>100))throw new RangeError("feature scores must be within [0, 100]");return values;});
  const labels=labelsRaw.map((v:any)=>{if(v!==0&&v!==1)throw new RangeError("labels must be binary 0/1 integers");return v;});const dates=datesRaw.map((v:any)=>dateValue(v,"observation_dates"));if(dates.some((v:Date,i:number)=>i<dates.length-1&&v.getTime()>=dates[i+1].getTime()))throw new RangeError("observation_dates must be strictly increasing");
  const split=Math.trunc(numberValue(data,"train_end_index"));if(split<8||split>rows.length-4)throw new RangeError("train_end_index must leave at least 8 training and 4 validation rows");const baseMap=data.base_weights;if(!baseMap||typeof baseMap!=="object"||Array.isArray(baseMap)||Object.keys(baseMap).sort().join("|")!==[...names].sort().join("|"))throw new RangeError("base_weights must match component_names");const base=names.map((n:string)=>nonnegative(baseMap,n));if(Math.abs(base.reduce((a:number,b:number)=>a+b,0)-1)>1e-9)throw new RangeError("base_weights must sum to 1");
  const learningRate=positive(data,"learning_rate"),ridge=nonnegative(data,"ridge_penalty"),slope=positive(data,"logit_slope"),iterations=Math.trunc(positive(data,"iterations"));if(iterations>5000)throw new RangeError("iterations must not exceed 5000");let weights=[...base],intercept=0;const trainRows=rows.slice(0,split),validationRows=rows.slice(split),trainLabels=labels.slice(0,split),validationLabels=labels.slice(split);
  for(let iteration=0;iteration<iterations;iteration++){const probabilities=trainRows.map((row:number[])=>sigmoid(intercept+slope*(row.reduce((s:number,v:number,j:number)=>s+weights[j]*v,0)/100-.5))),residuals=probabilities.map((p:number,i:number)=>p-trainLabels[i]),gradIntercept=residuals.reduce((a:number,b:number)=>a+b,0)/residuals.length,gradWeights=names.map((_:string,column:number)=>residuals.reduce((sum:number,residual:number,i:number)=>sum+residual*slope*trainRows[i][column]/100,0)/trainRows.length+ridge*(weights[column]-base[column]));intercept-=learningRate*gradIntercept;weights=projectSimplex(weights.map((v:number,i:number)=>v-learningRate*gradWeights[i]));}
  const baseTrain=brier(trainRows,trainLabels,base,0,slope),baseValidation=brier(validationRows,validationLabels,base,0,slope),candidateTrain=brier(trainRows,trainLabels,weights,intercept,slope),candidateValidation=brier(validationRows,validationLabels,weights,intercept,slope),improved=candidateValidation<=baseValidation,selected=improved?weights:base,selectedIntercept=improved?intercept:0,selectedValidation=improved?candidateValidation:baseValidation;
  return {state:improved?"calibrated-improved":"retain-base-weights",method:"nonnegative-simplex-logloss-v1",component_names:names,base_weights:Object.fromEntries(names.map((n:string,i:number)=>[n,base[i]])),candidate_weights:Object.fromEntries(names.map((n:string,i:number)=>[n,weights[i]])),selected_weights:Object.fromEntries(names.map((n:string,i:number)=>[n,selected[i]])),candidate_intercept:intercept,selected_intercept:selectedIntercept,base_train_brier:baseTrain,base_validation_brier:baseValidation,candidate_train_brier:candidateTrain,candidate_validation_brier:candidateValidation,selected_validation_brier:selectedValidation,weight_shift_l1:weights.reduce((s:number,v:number,i:number)=>s+Math.abs(v-base[i]),0),train_rows:trainRows.length,validation_rows:validationRows.length,coverage_ratio:1,reason:"temporal-holdout-governs-selection"};
}

const ROUTES:Record<string,[string,string,string[]]>={bank:["D18-F10-A01","basel-iii-teaching-v1",["cet1_ratio","leverage_ratio","lcr","nsfr","npl_ratio","provision_coverage_ratio","net_interest_margin","cost_income_ratio"]],insurance:["D18-F10-A02","solvency-ii-nonlife-teaching-v1",["scr_coverage_ratio","mcr_coverage_ratio","combined_ratio","adverse_reserve_development_ratio","tier1_own_funds_share","investment_concentration_ratio","return_on_equity"]],reit:["D18-F10-A03","nareit-equity-reit-teaching-v1",["nareit_ffo","recurring_capex","common_dividends","net_debt","ebitda_re","interest_expense","occupancy_ratio","same_store_noi_growth"]],utility:["D18-F10-A04","ferc-regulated-electric-teaching-v1",["allowed_roe","earned_roe","funds_from_operations","total_debt","debt_to_capital","ebit","interest_expense","capital_expenditure"]],"early-stage":["D18-F10-A05","early-stage-liquidity-teaching-v1",["unrestricted_cash","liquid_investments","monthly_net_cash_flows","near_term_obligations"]],cyclical:["D18-F10-A06","cyclical-midcycle-teaching-v1",["realized_price_history","unit_cost_history","volume_history","free_cash_flow_history","net_debt"]],"holding-company":["D18-F10-A07","holding-company-lookthrough-teaching-v1",["holdings","parent_cash","parent_debt","parent_interest_expense"]]};
function coverageDecision(data:RecordValue):RecordValue{
  requireFramework(data,"coverage-router-teaching-v1");const sector=textValue(data,"sector"),candidate=textValue(data,"candidate_framework"),asOf=dateValue(data.as_of,"as_of"),minimum=numberValue(data,"minimum_coverage"),maxAge=Math.trunc(positive(data,"max_age_days"));if(!(minimum>0&&minimum<=1))throw new RangeError("minimum_coverage must be in (0, 1]");if(!ROUTES[sector])return{state:"abstain",method:"sector-coverage-router-v1",selected_model:null,coverage_ratio:0,available_fields:[],missing_fields:[],stale_fields:[],future_fields:[],reasons:["unsupported-sector"],reason:"unsupported-sector"};
  const [model,expected,required]=ROUTES[sector];if(candidate!==expected)return{state:"abstain",method:"sector-coverage-router-v1",selected_model:null,coverage_ratio:0,available_fields:[],missing_fields:[...required],stale_fields:[],future_fields:[],reasons:["framework-mismatch"],reason:"framework-mismatch"};const facts=data.facts;if(!Array.isArray(facts))throw new TypeError("facts must be an array");const ledger:Record<string,any>={};for(const raw of facts){if(!raw||typeof raw!=="object"||Array.isArray(raw))throw new TypeError("each fact must be an object");const name=textValue(raw,"name");if(ledger[name])throw new RangeError("fact names must be unique");if(typeof raw.value_present!=="boolean")throw new TypeError("value_present must be boolean");ledger[name]=raw;}
  const available:string[]=[],missing:string[]=[],stale:string[]=[],future:string[]=[];for(const name of required){const fact=ledger[name];if(!fact||!fact.value_present){missing.push(name);continue;}const knowledge=dateValue(fact.knowledge_date,`${name}.knowledge_date`),periodEnd=dateValue(fact.period_end,`${name}.period_end`);if(knowledge.getTime()>asOf.getTime())future.push(name);else if((asOf.getTime()-periodEnd.getTime())/86400000>maxAge)stale.push(name);else available.push(name);}const coverage=available.length/required.length,reasons:string[]=[];if(future.length)reasons.push("future-evidence");if(stale.length)reasons.push("stale-evidence");if(missing.length)reasons.push("missing-required-fields");let state:string,selected:string|null;if(future.length){state="abstain";selected=null;}else if(coverage===1){state="supported";selected=model;}else if(coverage>=minimum){state="partial-review";selected=model;}else{state="abstain";selected=null;reasons.push("coverage-below-minimum");}if(!reasons.length)reasons.push("complete-current-coverage");return{state,method:"sector-coverage-router-v1",selected_model:selected,coverage_ratio:coverage,available_fields:available,missing_fields:missing,stale_fields:stale,future_fields:future,required_field_count:required.length,reasons,reason:reasons[0]};
}

export function calculate(topicId:string,data:RecordValue):RecordValue{
  if(!data||typeof data!=="object"||Array.isArray(data))throw new TypeError("input must be an object");
  const functions:Record<string,(value:RecordValue)=>RecordValue>={"D18-F10-A01":bankScore,"D18-F10-A02":insuranceScore,"D18-F10-A03":reitScore,"D18-F10-A04":utilityScore,"D18-F10-A05":earlyStageScore,"D18-F10-A06":cyclicalScore,"D18-F10-A07":holdingCompanyScore,"D18-F10-A08":calibrateWeights,"D18-F10-A09":coverageDecision};
  if(!functions[topicId])throw new RangeError(`unsupported topic ID: ${topicId}`);return functions[topicId](data);
}
Full-height labplaygroundOpen full screen