The proof, before you trust the answer

The single most common worry is that a page like this shows hand-picked, hardcoded numbers. So the headline figure is fetched live: when this page loads, your browser posts a real multi-year incentive stock option (ISO) case to the public production application programming interface (API) and reads back the answer. The optimized plan it returns, reported below as its after-tax Net Final Value (NFV), matches the value plotted in section 5 to the cent.

{
  "shares": 4000,
  "strike": 1,
  "fmv": 60,
  "filingStatus": "single",
  "ordinaryIncome": 300000,
  "stateCode": "CA",
  "carryforwardCredit": 0,
  "horizon": 2,
  "cashReturnRate": 0.04,
  "grantDate": "2023-01-01",
  "hasLeftCompany": false,
  "terminationDate": null,
  "expectedGrowth": 0.1,
  "volatility": 0.5
}

The calculators do not estimate the brackets and thresholds; they use the figures the IRS published for tax year 2026. Each value below is read directly from the production tax library and shown beside its source. A separate conformance test asserts every one of these against its published value and fails the build if a single digit drifts.

Primary source: IRS Rev. Proc. 2025-32 (including the One Big Beautiful Bill amendments), Form 6251, and the Internal Revenue Code sections cited above.

Matching the published constants is necessary but not sufficient; the arithmetic on top of them has to be right too. So we cross-check against PSL Tax-Calculator, an open-source federal individual-income tax model maintained independently of OptionsAhoy. Across 17 worked federal cases spanning ordinary income tax, long-term capital gains (the 0 / 15 / 20 percent stack plus the Net Investment Income Tax), and the Alternative Minimum Tax (AMT, including the ISO bargain element), the two engines agree to the cent.

All 17 cases agree to the cent. PSL Tax-Calculator 6.8.4. Each case gives taxcalc wages equal to OptionsAhoy's taxable income plus taxcalc's own 2026 standard deduction, so both engines reach the same taxable income and taxcalc applies the rules that key off the larger figure: the AMT adds the standard deduction back (Form 6251 line 2a) and the NIIT tests AGI. The itemizer case feeds its state and local taxes as e18400 instead. ISO bargain element fed as cmbtp; marginal capital-gains figure is the iitax delta (taxcalc iitax already includes NIIT). All 2026-law parameters are taxcalc defaults, which match the published Rev. Proc. 2025-32 / OBBBA figures exactly.

Correction, October 2026: this check previously ran PSL Tax-Calculator 6.6.2 with its standard deduction set to zero. That version also left the standard deduction out of AMT income, so both engines agreed on AMT figures that were too low: by $4,508 to $6,762 for the single filers in this table and $9,016 to $13,524 for the joint filers. Form 6251 adds the standard deduction back (line 2a). The current PSL Tax-Calculator does, and so does OptionsAhoy.

Federal tax is only part of an equity decision; for most holders the state bill is the larger swing. So the state rate schedules are cross-checked the same way, against OpenTaxSolver, an open-source tax program that encodes each state's official rate schedule and is maintained independently of OptionsAhoy. Across the five states that carry the most equity compensation, CA, NY, NJ, PA, MA, the marginal tax OptionsAhoy computes on a sale matches the independent engine to the cent.

All 16 cases agree to the cent. OpenTaxSolver 2025_23.06, tax year 2025. The comparison is at the state rate schedule (the marginal bracket math); each schedule is also confirmed against the state's official publication.

A rate schedule is only part of a state return. Each state starts from its own income figure and allows its own deductions, exemptions and credits, several of which phase out as income rises. So the whole state return is cross-checked against PolicyEngine US, an open-source model of every state's income tax maintained independently of OptionsAhoy. Across all 42 states with a wage income tax and 504 cases (single, joint and head-of-household filers from $100,000 to $1.2 million of taxable income, with and without capital gains), 360 agree to within $1. The other 144 were each traced to their cause in the state's own publications:

• AL: PolicyEngine's Alabama taxable income is lower than Form 40 gives. At $116,100 of AGI, Form 40 line 12 allows $16,712 of federal income tax, leaving $95,388 of taxable income after the $2,500 standard deduction and $1,500 exemption; PolicyEngine reaches $89,006. OptionsAhoy follows the form.
• AR: Rounding in the 2026 table, under $5.
• AZ: PolicyEngine 2.29.13 keeps Arizona's 2025 standard deduction ($15,750 single). A.R.S. 43-1041 indexes it the way the federal amount is indexed: $16,100 for 2026.
• CA: The Franchise Tax Board published California's 2026 rate schedules in October 2026, and OptionsAhoy uses them; PolicyEngine 2.29.13 projects them from 2025.
• DC: PolicyEngine indexes the District's 2026 standard deduction. Under D.C. Code 47-1801.04 the first adjustment comes in 2027, so 2026 stays $15,000 single, $30,000 joint.
• ID: Idaho's $10 Permanent Building Fund tax, which PolicyEngine leaves out of the compared figure, net of its projection of the 2026 zero-rate amount.
• MA: PolicyEngine projects the 2026 surtax threshold at $1,107,700; Massachusetts set it at $1,107,750. $2.
• NM: New Mexico's $2,500 capital gain deduction applies only to net long-term gain (NMSA 7-2-34); PolicyEngine also applies it to a short-term gain.
• OR: PolicyEngine 2.29.13 estimates Oregon's 2026 indexed amounts; OptionsAhoy uses the amounts in Oregon's 2026 withholding formulas.
• VT: Vermont has not published its 2026 standard deduction. OptionsAhoy indexes the statutory amounts with the law's own formula, which reproduces every figure Vermont has published; PolicyEngine projects them differently.
• ME: Under $10; rounding or 2026 parameters not yet published. Not traced individually.
• MI: Under $10; rounding or 2026 parameters not yet published. Not traced individually.
• MN: Under $10; rounding or 2026 parameters not yet published. Not traced individually.
• NE: Under $10; rounding or 2026 parameters not yet published. Not traced individually.
• NY: Rounding in the recapture computation, $1.
• WI: Under $10; the sliding-scale deduction is computed from the statute, not the rounded table. Not traced individually.

No unexplained differences. PolicyEngine US 2.29.13, tax year 2026. W-2 earner on the standard deduction (itemizing off), wages = federal taxable income + the 2026 federal standard deduction, one earner on a joint return, no dependents; compares state income tax before refundable credits, local taxes excluded.

For a deliberately small two-year case, the entire decision is one number: how many of the 4,000 shares to exercise in year one (the rest fall in year two). That lets us brute-force every possibility at single-share resolution and plot the after-tax Net Final Value (NFV). The schedule the optimizer recommends sits on the peak of that curve, above both naive baselines.

Optimizer pick: 1,954 shares in year one, $123,560.45. That is identical to the single-share brute-force maximum, and $20,944.59 ahead of exercising everything now. On real multi-year grants the space is far larger than a person can enumerate by hand, which is the entire point of the tool.

Beyond specific cases, the engine is checked against properties that have to be true for any input. These run on every build:

• The optimized plan is never worse than exercising everything now or splitting evenly.
• Exercising more shares never reduces the gross gain before tax.
• Alternative Minimum Tax credit recovered never exceeds the credit earned.
• At the crossover share count, tentative minimum tax equals the regular tax, by construction.
• Out-of-range or malformed inputs return a clear error with the offending field named, never a wrong number.

None of this requires trusting a screenshot. The calculation engine is open source under the MIT license at github.com/AlvisoOculus/optionsahoy-mcp. The independent cross-checks above are produced by committed scripts that anyone can run against PSL Tax-Calculator (federal), OpenTaxSolver (state rate schedules) and PolicyEngine US (whole state returns); the federal and state agreement, the constant conformance, and the optimizer ground-truth check are all enforced by the automated test suite, so the claims on this page cannot silently go stale. You can also re-run the live API call yourself with the payload shown in section 1.

References: IRS Rev. Proc. 2025-32 · IRS Form 6251 · PSL Tax-Calculator · methodology