Mini test 001 / Buy at close, sell at open / Micron
Can trading Micron overnight really turn $1,000 into $1.67 billion?
A widely shared backtest claimed that buying Micron at every market close, selling it at the next open, and reinvesting the whole account would turn $1,000 into about $1.67 billion. We checked what a real account might capture—and whether it would beat simply holding Micron stock, whose ticker is MU.
The overnight trade made money. Holding MU made far more.
The rule sat in cash during the day, so it was already behind before costs. Trading on 495 nights then cut $5,912 to $3,603. It did reduce the worst fall—from 43% to 29%—but not enough to make it the better investment.
01 / The original claim
How does the calculation get so large?
From 1989 through 2026, it compounds every published close-to-open move in MU. It assumes the account receives the exact official closing price, receives the exact opening price the next morning, and pays nothing to trade. Reinvesting that historical gap across 9,477 nights turns $1,000 into $1.67 billion.
The arithmetic is real. It is not a record of prices an actual account received.
02 / The obvious cost question
Is our assumed cost realistic?
The proposed automated version used fractional-share orders at the broker Public. We assumed $1 of total cost each night for every $1,000 traded. Public lists no commission for regular stock trades; even its paid routes cost only $0.003 a share. Most of our $1 is therefore a guess for the bid–ask spread and the price moving before a fractional order fills—not a listed fee.
We do not have the old bids, asks, or Public fills needed to know the true number. It may be lower on quiet nights and higher at volatile openings. Public also says fractional dollar orders are market orders and may receive several fills. Read Public's fees and its fractional-order disclosure.
For the 37-year headline, cost decides everything: it must stay below about 0.083% every night to beat holding MU. For the two-year minute-price comparison, it does not change the choice. Even at zero cost, holding MU still won.
03 / What this establishes
What can we really say?
MU really did tend to rise between the published close and open. In the two recent years where we have minute prices, a version using the last trade near the close and the last trade in the opening minute also made money after our assumed cost.
We cannot say Public can capture that return. We cannot say it beats holding MU. We cannot say the effect works across stocks. And we cannot treat the billion-dollar result as money an account could have made.
So the honest conclusion is: an interesting MU price pattern, not a proven strategy, and no reason to replace holding MU with it.
04 / What would come next
What would make this a real strategy test?
The rule would have to become generic rather than depend on one stock. Before seeing the answer, we would define which stocks it may trade, when it buys and sells, the costs, what it must beat, and which later period stays untouched for the final test.
Until that fuller process is complete, this belongs in Insights as a useful check of a claim—not in the strategy library.
The full receipt
What exactly did we test?
The recent account test
From 21 August 2024 through 20 August 2026, the account bought MU with its whole balance using the last recorded trade in the 3:59 p.m. minute. It sold using the last trade in the next opening minute, then reinvested the whole balance the next night. The assumed total cost was 0.10% per night.
| Choice | Ending $1,000 | Worst fall |
|---|---|---|
| Hold MU | $8,685 | 43% |
| Overnight, no cost | $5,912 | 26% |
| Overnight, 0.10% nightly cost | $3,603 | 29% |
| Hold QQQ | $1,491 | 23% |
The long published-price calculation
The 1989–2026 calculation compounds 9,477 published close-to-open moves. With no cost, $1,000 becomes $1.67 billion. At 0.10% total cost each night it becomes $127,299, while holding MU becomes $620,798. At 0.20% nightly cost, the overnight calculation ends with $9.68.
These are adjusted daily prices, not a record of automated fractional-order fills.
Sources and remaining limits
The idea came from this X post. The test is frozen as MU-OVERNIGHT-EXECUTION-001. Daily adjusted prices came from Yahoo Finance; recent minute trades came from Massive. We could download only two years of minute prices, and minute trades do not show the historical bid, ask, or Public's actual fills.
This is educational research, not personal investment advice or a recommendation to trade MU.
The next test
See the next claim tested—not merely repeated.
Subscribe, then optionally tell us which investment claim or strategy you want checked next.