Own one stock: the biggest company in the S&P 500. Check every three months. If a different company has been on top for two checks in a row, sell what you hold and buy that one. Otherwise do nothing — which is what almost every check comes to. Every figure below assumes you put in $10,000 at the start and reinvested the dividends.
| Check | Who is #1 | Action | Why |
|---|---|---|---|
| Q1 | Apple | HOLD | Apple was #1 last quarter too. Nothing to do. |
| Q2 | Microsoft | WAIT | New name at the top. You still own Apple. One quarter is not enough. |
| Q3 | Microsoft | SWITCH | Two checks in a row. Sell Apple, put everything into Microsoft. |
| Q4 | Amazon | WAIT | New name again. You still own Microsoft. |
| Q5 | Microsoft | HOLD | Amazon did not stay #1, so it never earned the switch. You still own Microsoft — and you never traded on its one good quarter. |
Everything above this line is a reconstruction — what the rule would have returned, worked out afterwards from historical prices. This part is not. It is one real Roth IRA, following the rule forward from the day it was opened, priced at the last close on record. It will look nothing like the twenty-year figures for a long time, and that is the honest version of what starting today looks like.
Every quarter, check who holds the #1 spot in the S&P 500. If a new company has held it for two quarters running, sell everything and buy that one. Otherwise, hold.
Not investment advice. Sandpone is a public record of a mechanical rule set, published for discussion and curiosity. Nothing here is a recommendation to buy or sell any security.
Backtested, not traded — except where it says otherwise. The twenty-year record, the chart, the simulator and every headline figure are reconstructed from historical data with the benefit of hindsight about which company held the top spot. No money was invested on those terms. The only exception is the "Real money" section, which is one actual Roth IRA opened after this site existed and priced from daily closes; it is labelled as such and its result should not be read as the strategy's twenty-year record. Backtested returns are pre-tax and ignore commissions, spreads, and the capital gains due on every switch outside a sheltered account.
Provenance. Prices and dividends from Yahoo Finance; inflation from BLS series CUUR0000SA0; share counts for the leaderboard from SEC XBRL, which begins in 2009. Every figure is computed by ingest.py and leaderboard.py in the project repository and stored as fetched — the page renders those records and calculates nothing of its own. Nothing here is interpolated: a missing input renders as a gap.
What "#1" means here. The biggest company in the S&P 500, as the index itself ranks them — by float-adjusted market value, meaning only the shares the public can actually buy. That number is not an opinion: it is read from the holdings file of iShares IVV, the fund that tracks the index, at each quarter end. Every switch on this page was confirmed at both of the two quarter ends that triggered it, and both are re-checkable with index_leader.py in the project repository.
Why the record starts in September 2006. That is the earliest holdings snapshot BlackRock publishes for IVV. It is a floor imposed by the data, not a start date chosen for how the numbers look. Two quarters (2017 Q1 and Q2) have no snapshot at all; Apple led both neighbouring quarters by a wide margin, and a missing quarter can never count toward a confirmation, so the gap cannot hide a switch.
The two-quarter rule was chosen with hindsight. It was tested against eleven alternatives on this same twenty-year record and beat rotating-on-the-instant from all 58 possible starting quarters. But those 58 runs overlap and share a single market history, and the advantage rests largely on four bad switches avoided. That is evidence, not proof. The full comparison is backtest.py in the repository; run it and disagree.
Leaderboard coverage. The standings table ranks by full market cap within a fixed candidate list of large-cap index members, not all 500. Companies whose share count cannot be resolved from unsegmented SEC XBRL — some multi-class issuers report only one class — are excluded outright rather than ranked on a partial count.