DERIVATIVES · MODEL GOVERNANCE

Last reviewed: 7 October 2026 · Regression reference: 28 cases

Calculator Methodology & Validation

How the RegTech Nexus AI derivatives calculators should be interpreted, tested and governed before any professional or institutional reliance.

1 · PURPOSE

Analytical models, not production engines

The calculators are client-side educational and analytical tools. They demonstrate recognized financial-engineering relationships using user-supplied or illustrative inputs. They are not exchange, CCP, bank or institutional valuation engines and do not establish regulatory compliance.

Boundary: A standard formula is not, by itself, evidence that an implementation is independently validated. Material outputs should be recomputed and verified against authoritative product conventions before professional reliance.
2 · MODEL FAMILIES

Primary calculation frameworks

Black-Scholes

European-style option analysis under stated assumptions. The public implementation uses a high-precision browser-compatible normal CDF, with Actual/365 maturity conversion; inputs and compounding conventions must be interpreted exactly as labelled.

Black-76

European-style options on futures/forwards with positive forward/futures and strike inputs, continuous compounding and Actual/365 maturity conversion. Theta is reported per calendar day and includes the r × option-value discounting term. It does not cover all commodity option exercise styles or negative-price environments.

Cost of carry

Illustrative forward/futures relationships using financing, storage and convenience-yield assumptions. The Storage & Carry calculator uses simple-interest financing on the stated flat commodity value and does not compound or finance accumulated storage/handling costs. Actual market pricing may include basis, transport, quality, location and contract effects.

Risk / hedge measures

DV01, Greeks and minimum-variance hedge ratios are simplified analytical sensitivities, not substitutes for portfolio risk engines.

3 · VALIDATION STANDARD

Recommended regression evidence

  1. Record calculator name, model, version, inputs, units and assumptions.
  2. Independently recompute the expected result using a trusted reference implementation.
  3. Compare expected versus actual output using a defined numerical tolerance.
  4. Test boundary conditions: zero/negative/blank inputs, extreme values, maturity boundaries and invalid correlations.
  5. Test economically meaningful identities where applicable, such as option parity, long/short sign conventions and put/call directionality.
  6. Repeat across supported browsers and retain evidence of the test environment.
  7. Retain the test vector, expected output, actual output, tolerance, date and reviewer.

Evidence status: Fact · Observation · Inference · Assumption · Unverified · Recommendation.

4 · KNOWN PRODUCTION GAPS

What the public calculators do not model

Market data

No live institutional curves, volatility surfaces, market-data quality controls or independent price verification.

Valuation complexity

No general-purpose multi-curve, stochastic-volatility, Monte Carlo or full XVA framework.

Margin

Indicative scenarios are not CCP SPAN/CORE, ISDA SIMM or other approved production margin engines.

Governance

Public calculators should not be treated as independently validated models, regulatory capital engines, client-pricing engines or automated trading systems.

5 · ASSURANCE RULE

Evidence before conclusion

Use the RegTech Nexus audit sequence: DISCOVER → INSPECT → VERIFY → TEST → CHALLENGE → ASSESS → RISK-RATE → REMEDIATE → RETEST → ASSURE.

Where the evidence needed to support a conclusion is unavailable, use the control statement: INSUFFICIENT EVIDENCE — VERIFICATION REQUIRED.

6 · REGRESSION REFERENCE VECTOR

28 reference cases · 7 October 2026

These reference cases were independently recomputed during the validation review. They are regression controls for future code changes, not evidence of external model certification. The core and commodity Greeks interfaces now explicitly support call/put and long/short direction; the published sample remains long call with zero dividend yield so the historical reference vector remains comparable.

Core suite · 14 cases

Futures P&L 10,000 · FX forward 111.08374384 · Black-Scholes 2.44578940 · option payoff 700 · IRS -12,500 · FRA 12,330.4562 · DV01 4,500 · swap rate 5.9052059% · option Greeks reference vector passed · basis 2 · hedge ratio 1.02 / notional 10.2m · indicative margin 800,000 · breakeven 106 · collateral interest 16,666.6667.

Commodity suite · 14 cases

Futures P&L 15,000 · carry 77.25233599 · basis -2 · contract conversion 210,000 · Black-76 3.15921673 · Greeks price 4.39857160 / delta 0.52319734 / theta -0.02378881 per day · calendar spread position value 6,000 · intercommodity spread 8 · crack 35/barrel · crush 0.9418 (CME-aligned soybean sample) · hedge ratio 0.765 · storage direct 16,800 / financing 18,750 · swap 30,000 · margin stress 660,000 / adverse loss 900,000.

Regression rule: retain input vector, units, assumptions, expected value, tolerance, actual value, environment and reviewer/date. Rerun affected cases after material changes. A mathematical pass does not establish production suitability, CCP approval, regulatory compliance or independent model validation. Export reports are evidence records; actual values must be retained separately from reference values and compared using the documented tolerance.
Important: This methodology is an internal/public analytical governance framework, not a legal, accounting, investment or regulatory opinion. Applicable requirements depend on jurisdiction, entity status, product, transaction and authoritative primary sources.