How the record is measured

One portfolio, two accounts

I manage two accounts held by my household as one portfolio, the composite. Transfers between the two are internal and net to zero. Two things held in the accounts stay out of it.

The carve-out. Long-dated US Treasury bonds: those maturing in 2044 and 2048, with 3⅜% coupons, held to maturity in both accounts, and one maturing in 2053, held during 2024. I bought them with new money from outside the portfolio, and they sit outside the active strategy. Their market value and accrued interest stay out of the composite, and the coupons they pay into the accounts count as money coming in, never as return. A carve-out bond bought or sold with the composite’s cash counts as a flow too.

Savings set aside. Savings kept in one of the accounts, since May 2026 in a short-dated Treasury note. They aren’t part of the portfolio: the part of the note held as savings stays out of the composite with its accrued interest, money moved into savings counts as money leaving the portfolio, and money taken back counts as money coming in.

Other than the savings, Treasury bills, short-dated Treasury notes and short-term Treasury funds stay inside the composite: they are its cash.

The daily return

Each day’s value, V, is the net asset value of both accounts at the close, minus the carve-out and the savings set aside. The flows, F, are money entering or leaving the household, signed (withdrawals are negative), plus the carve-out’s coupons and any change in the savings set aside.

Daily returns are linked geometrically into months and years, so the return is time-weighted: money coming in or going out changes the size of the portfolio, not its return. A calendar period ends on its last valuation.

Valuation

Every trading day since 15 Sep 2023, from the broker’s daily statements of both accounts

Costs

Net of commissions, fees, margin and borrowing interest, and dividend withholding tax; gross of my personal income tax; no management fee

Currency

US dollars, with no conversion

Inception

15 Sep 2023, the first recorded value; full time since January 2024, marked as an event, not a restart

Exposure

Each comparison needs to know how much of the portfolio was in the market each day. That share is my exposure, w: the exposure of every position added up, divided by V.

Stocks and ETFs count at market value. An option counts for the shares it behaves like: contracts × 100 × delta × the share price, with the delta my broker’s platform shows just before the close. Those deltas weren’t kept before 23 Oct 2026, so for earlier days each option’s delta is estimated from its closing price; the record marks which days are which. A short put adds exposure, and a short call subtracts it. A synthetic long, a long call and a short put on the same stock with the same strike and expiry, counts as the shares themselves, because by put-call parity the pair moves one for one with a share. Treasury bills, short-term Treasury funds and bonds count as cash. For the comparisons, w is used as measured, floored at zero. It is never capped: a day above 150% is flagged for a check.

It’s the conversion EU fund rules use for the commitment approach (UCITS and AIFMD) and the one behind the delta-adjusted figures in US fund rules (SEC Rule 18f-4) and IOSCO’s leverage framework. This is a private portfolio, not a regulated fund: the rules are cited for the method, not as a claim of compliance.

Delta measures exposure to small moves. A short put’s delta grows as the stock falls, so in a gap the book loses more than delta alone says. The statistics pack shows, next to the delta exposure, the loss if every stock fell 20% overnight.

The whole record follows this rule: it is rebuilt from the broker’s daily statements back to 15 Sep 2023.

Two funds I could hold instead

I benchmark against what I would own if I stopped. JEPI and BALI are the two off-the-shelf funds closest to what I do: large US companies, income from options, less volatility than the market. I could hold them instead without managing anything. If over a full cycle I can’t beat them, after matching how much of the portfolio was actually invested, the honest conclusion is to buy them and stop. That’s what the comparison is for: it’s the test that could end the strategy.

Benchmarking against peers may seem to contradict an absolute return mandate, but market conditions and the stage of the credit and economic cycle have to be taken into account. In the low stages of the cycle, the minimum absolute return may not be met, but then neither comparable ETF would meet it. Missing the absolute return target in a given year is therefore not enough to trigger the kill rule: for that I need comparable passive strategies with similar goals, and these are JEPI and BALI.

JEPI is the JPMorgan Equity Premium Income ETF, actively managed, listed since May 2020. BALI is the iShares U.S. Large Cap Premium Income Active ETF, also actively managed, listed since 26 Sep 2023. Their daily returns are total returns at the market price: the closing price plus distributions, reinvested on the day the fund goes ex-distribution. That is what a holder earns; it differs from the return on net asset value by the change in the fund’s small premium or discount. Both funds pay out a lot, so price return alone would understate them. The comparisons run from 15 Sep 2023, the record’s first day; BALI joins on 28 Sep 2023, the first close in its price data, from my record’s level that day. For context only, the record also shows the S&P 500, with its dividends, held at my exposure the same way, and 3-month Treasury bills at the Federal Reserve’s daily rate, accrued daily. The index is not a target: my mandate is absolute return, and a portfolio fully in stocks isn’t one I would hold, because the family’s spending needs a source of cash other than selling assets.

Two comparisons

I’m rarely fully invested. Each day the comparison holds the same share of capital in the fund as I held in equities at the previous close, with options counted by their delta at the share price, and the rest in Treasury bills. A portfolio 30% invested is compared with a fund position 30% invested. A second version holds the fund at my average exposure, which shows whether changing mine added anything.

Both comparisons match capital, not risk. JEPI and BALI carry less equity risk than their capital weight suggests, which flatters my portfolio in rallies and penalizes it in sell-offs.

I choose “return at the same exposure” over the standard “risk-adjusted” return because of how risk is defined in most contexts: as volatility. When options are the main tool for expressing a market view, risk-adjusted measures such as the Sharpe ratio are unreliable, because the distribution of returns is no longer normal, and so are the readings they produce. Why? Implied volatility is the lifeblood of the options market. I can tolerate more volatility than a traditional investor, because higher volatility also increases the extrinsic value of options, which is an asset for the option seller. Besides, if you can tolerate a wide band of price volatility, managing to risk-adjusted measures only lowers your expected return, which is undesirable too.

The kill rule

Every year at the December close, starting in December 2026, I compare my return over the previous 36 months with JEPI’s and BALI’s on two measures: each fund held at my exposure day by day, and each fund held at my average exposure for the period. If I trail both funds on both measures, the rule has triggered.

Both measures have to agree, so the reading is one consistent test and the measure can’t be picked after the fact. The dates are fixed for the same reason: the test runs on its date, not when a result looks good or bad.

If it triggers

These steps were set before any test, so the result can’t shape them:

1. Find where the shortfall came from. It is concentrated if it disappears once the two worst quarters, or the three positions that cost the most, are left out; otherwise it is structural. A concentrated shortfall usually carries a lesson, such as a costly error, a weight too large, an exit too slow or an entry too fast, and a new guardrail can prevent it. A structural one is harder: I have to separate luck from skill and name the causes that run through the whole window. If I can’t name them, I move the portfolio into the funds.
2. Try the changes for at most four quarters, until the next December test, whatever the market does. The trial is judged on its own year, not on the 36-month window, which a year later is still two-thirds old results: it passes if I trail neither fund, held at my exposure, by more than 0.75 percentage points. A gap that small, either way, is noise in a single year, not a verdict.
3. Three rounds at most. If a trial fails, back to step 1. If the third fails, I move the portfolio into JEPI and BALI and step aside.

Currency

Returns are in US dollars, the currency the portfolio earns in, and so are the benchmarks. The family spends in euros: about 5% of capital a year is converted when it’s needed, at that day’s rate, so no single exchange rate decides the outcome. I’d hedge only if I expected the dollar to lose its role for good, and I don’t. I invest in the US because I manage risk with listed options, and no other market comes close in depth and liquidity.

What I publish, and what I don’t

The weekly figures are provisional: until a quarter is frozen, a figure in it can still be corrected. Once frozen, a change is a new version of the file, with a dated note (see Revisions).

To ask for any of it, write from a work address with your name, firm and purpose. I reply personally, with a PDF stamped “Prepared for [name, firm], [date]”.

How to check a file

The record is kept daily. At each quarter-end, that quarter’s daily returns and exposure are frozen in one file, record-YYYY-Qn.csv, in percentages only; the first file, frozen at launch, holds the whole history since 15 Sep 2023. Each file is fingerprinted. A fingerprint (SHA-256) is a 64-character code computed from a file’s bytes: change one digit and the code changes completely, and the code reveals nothing about the contents. Each file’s header names the method version and the previous file’s fingerprint, so the files form a chain, and an edit to any past quarter breaks it.

The fingerprints go on the ledger and on the Record page, and both are saved to the Wayback Machine the same day. The date comes from that third party, not from me.

To compute a fingerprint:

Windows, PowerShell   Get-FileHash record-2026-Q4.csv -Algorithm SHA256
macOS, Terminal       shasum -a 256 record-2026-Q4.csv

A matching code shows the file hasn’t changed since the day its fingerprint became public. It doesn’t show the numbers are right; the statements do that. And it never shows when a file was written, only that it existed by the date of the archive. The research notes work the same way: each note lists the fingerprint of every PDF version, so anyone can download a PDF, compute its code and compare it with the archived copy of the note’s page.

Limits

• The comparisons match capital, not risk.
• The funds are measured at their market price, not their net asset value.
• Before 23 Oct 2026, the exposure rests on deltas estimated from closing prices. On 152 trading days in 2023 and 2024 it also leaves out options on four stocks whose past prices I couldn’t find.
• The savings are taken out of the composite from May 2026, when they went into the Treasury note. Savings kept earlier as US dollars have no record, and they were negligible against the portfolio.
• Notes written before the site went live carry the date I wrote them, which only I can vouch for; their public date is launch day.

Revisions

A change to a frozen record is a new version, with a dated note saying what changed and by how much; the old version and its fingerprint stay on the ledger. A change to this method is a new, dated version of this page, with the reason, and the old one stays archived.

Changes to the method

• None since version 1.