Trend-gated mega-cap equity, tested until it broke

A rule that holds the fifteen most-traded S&P 500 companies, replaces them monthly, never trims a winner, scales exposure to a volatility target, and moves to cash when the market falls below its long-term average. Every number here comes from a point-in-time backtest whose arithmetic reconciles to raw source files within 0.05 basis points.

Growth

Logarithmic scale · monthly marks

Tap the chart to read exact values

Drawdown from peak

Net exposure — the gate and the volatility target at work

The same rule across four eras

The advantage is not uniform. It was overwhelming through the lost decade and has been absent since 2018. Both are shown.

Era	Book	CAGR	Max DD	Sharpe	Calmar	Worst year	becomes

Swipe the table sideways to see all columns

Section 02

Every month, every year, every drawdown

Nothing removed, nothing smoothed. The bad years are the point of the record, not an omission from it.

Monthly returns

per cent

Swipe sideways for the full year

Calendar years against QQQ

Annual return · strategy versus the index

Drawdowns past 15%

Dated from peak, through trough, to the day the previous high was reclaimed. The bar shows total time submerged.

Peak	Trough	Recovered	Depth	Months	Duration

Section 03

Position history

Every holding the rule opened and closed across twenty-six years, from the simulation. These are backtest positions — no capital was ever committed.

Currently indicated book

What the frozen rule points at on the most recent data. Not a recommendation, and not a position anyone holds.

Symbol	Target weight	Last price	at capital

All closed and open positions

Symbol	Entered	Exited	Held (days)	Weight at entry	Return

Return is measured over the whole holding period, not as a contribution to portfolio return — a position entered at 5% and held for twenty years is not twenty years of 5% exposure. A profit-factor figure is deliberately not published here: computed on these numbers it exceeds 700, which would be meaningless rather than impressive.

Section 04

Historical replay

Pick an amount and a starting month, and this replays what the backtest says would have happened between then and the end of the data. It is a record of the past, not a projection.

Starting capital (USD)

Invested at the end of

Sold at the end of

Added every month (optional)

Replay

Read this before believing the number above. The same rule also spent months below its previous high, and lost in its worst calendar year. Choosing a start month with hindsight is the easiest way to mislead yourself — try January 2000, then January 2010, then January 2018.

Capital goes in at the end of the month selected, so a replay from the very first month skips that month and will not match the headline exactly. The drawdown shown uses month-end marks, shallower than the daily figure of quoted elsewhere.

Section 05

What was verified, and what was not

The reason to trust any of the preceding numbers is not that they are large. It is that each one has a file behind it, and that the checks below were built to fail loudly.

Verification ledger

Check	Result	Status	Evidence

Seven defects were found by these checks — two in the engine, five in the test harness itself, including a walk-forward whose candidate pool had been contaminated with full-sample information. Each is listed in AUDIT_EVIDENCE.md. A battery that never fails is a battery that is not looking.

Not proven

Given the same weight as the section above, because a research record that buries its gaps is advertising.