November 16, 2026 — prices through November 13, 2026
This experiment investigates buying the dip in the Magnificent Seven stocks since the group acquired its name in May 2023. Buying stocks already at least 25% below their previous records at inception, then buying each fresh qualifying decline, generated 15 purchases. Holding every purchase through November 13, 2026, with dividends reinvested, turned $15,000 into approximately $23,527, producing an annualized money-weighted return of 28.04%.
A stricter version that waits for fresh declines generated 12 purchases, turning $12,000 into approximately $16,445, with an annualized return of 28.11%. Including the earlier purchases increased dollar profits substantially but left the annualized return almost unchanged.
Question 1: When does the experiment begin, and what triggers a purchase?
Bank of America strategist Michael Hartnett’s May 19, 2023 research note identifies Apple, Microsoft, Google, Amazon, NVIDIA, Meta and Tesla as the Magnificent Seven. The experiment begins on May 22, the next trading session, and retains those seven companies throughout. Alphabet is represented once, through its GOOGL Class A shares.
The rule buys $1,000 when a stock first closes at least 25% below its highest previous closing price, adjusted for stock splits. The purchase occurs at the following trading day’s opening price. Dividends are included in investment returns but do not change the price series used to identify buying signals.
Each decline permits only one purchase. Repeated crossings of the 25% boundary generate no additional $1,000 purchases. The rule resets when the stock regains its previous record closing price; a subsequent qualifying decline then permits another $1,000 purchase.
There are no sales or profit taking.
Dividends are reinvested without additional contributions, and fractional shares are allowed.
Previous records include prices before May 2023. The relevant records for Amazon, Meta and Tesla were set in 2021. All seven had crossed the 25% threshold before inception, but only Amazon, Meta and Tesla remained at least 25% below their records at the May 19, 2023 close.
The main scenario buys $1,000 of each of those three stocks at the May 22 opening. Their earlier declines count as one purchase each, and each stock must regain its previous record before another purchase is permitted. Apple, Microsoft, Alphabet and NVIDIA had already recovered above the 25% boundary, so they do not receive initial purchases.
The comparison scenario excludes every pre-inception signal and waits for each stock to regain its record before a fresh decline can trigger a purchase. Its first purchase occurs in August 2024. Both scenarios subsequently make the same 12 purchases; the only difference is the three initial positions.
Question 2: What were the returns, and how much risk remained after buying?
Including stocks already below the threshold. The main scenario makes three purchases each in Meta, NVIDIA and Tesla; two each in Amazon and Alphabet; and one each in Apple and Microsoft. Its $15,000 grows to approximately $23,527, a gain of $8,527, or 56.8% of contributions. Its annualized money-weighted return is 28.04%.
Waiting for fresh declines. The comparison scenario makes three purchases in NVIDIA; two each in Alphabet, Meta and Tesla; and one each in Apple, Amazon and Microsoft. Its $12,000 grows to approximately $16,445, a gain of $4,445, or 37.0% of contributions, with an annualized money-weighted return of 28.11%.
Fourteen of the 15 purchases in the main scenario show gains; 11 of the 12 in the fresh-decline scenario show gains. The latest NVIDIA purchase is the sole loss in both. NVIDIA’s three purchases combined have gained 32.5%. It is a relatively recent purchase, which is still falling. The jury is still out.
The annualized figures use XIRR, which accounts for the amount and timing of each contribution and the final portfolio value. They assume a fresh $1,000 is available at each purchase, rather than committing all $15,000 or $12,000 at inception. Gains as a percentage of contributions are cumulative figures, not annualized returns.
How do the results compare with the S&P 500? For each scenario, the benchmark invests the same $1,000 amounts on the same dates in SPY, the SPDR S&P 500 ETF, with distributions reinvested.
For the 15-purchase scenario, the matched SPY investments grow to approximately $21,085, a 40.6% gain on contributions, with an annualized money-weighted return of 21.01%. The Magnificent Seven purchases finish approximately $2,443 ahead.
For the 12-purchase scenario, the matched SPY investments grow to approximately $15,439, a 28.7% gain on contributions, with an annualized money-weighted return of 22.12%. The Magnificent Seven purchases finish approximately $1,006 ahead.
Both sides include reinvested dividends, and SPY’s returns reflect fund expenses. Both exclude taxes, trading commissions and interest on money awaiting investment. These comparisons measure the choice of investment on the purchase dates. They do not establish that waiting for declines beat investing available money earlier.
Buying after a large decline also left considerable risk. The February 2025 Tesla purchase subsequently fell 35.8% below its entry price on a closing basis before recovering. The March 2026 Tesla purchase fell 18.5%, and the March 2025 NVIDIA purchase fell 15.2%. These price-based risk measures compare purchase prices with the lowest subsequent closes through November 13; intraday losses could have been larger.
The portfolios change composition as purchases accumulate. The main scenario begins with three stocks, while the fresh-decline scenario begins with a single NVIDIA position. Both remain concentrated in seven large technology and technology-related companies. Neither return should be interpreted as the performance of a broadly diversified portfolio.
Question 3: Does the evidence support buying NVIDIA now?
The rule already triggered on November 9, 2026. A $1,000 purchase at the November 10 opening price of approximately $173.97 would now be worth $936, a loss of 6.4%. There was no dividend ex-date for that purchase during the measured interval. Remaining below the threshold does not trigger another $1,000 purchase.
The scandal concerns reported affairs involving Jensen Huang, including an alleged relationship with a woman suspected of working for Chinese intelligence, raising questions about corporate governance and access to restricted chips. Reuters reported that the Justice Department had not determined whether she was a spy, while NVIDIA’s board had launched an investigation.
NVIDIA’s November 12 announcement confirms Huang’s resignation and the appointment of Chief Financial Officer Colette Kress as interim chief executive. The company said it expected its operations, customer commitments and product roadmap to continue without interruption.
The Wall Street Journal has examined Kress’s concerns about Huang’s dealmaking and NVIDIA’s financing of customers unable to afford its chips. Reporting on the Journal’s subsequent coverage says startups expect less financing under Kress. A pullback could reduce NVIDIA’s financial exposure while weakening customers’ ability to buy its products.
Against these risks, NVIDIA’s valuation has fallen substantially. Yahoo Finance’s valuation statistics, retrieved November 16, show a forward price-to-earnings ratio of approximately 10.3, compared with 22.9 at the end of July 2026 and 31.5 at the end of October 2025. These figures suggest the price already reflects some deterioration in the outlook.
The investment question is whether the lower price adequately compensates for leadership disruption, financing changes and possible restrictions on future business. Previous recoveries cannot answer that question.
The historical results are encouraging, but the main scenario contains only 15 purchases, including three made on the same day and several others clustered around common selloffs. They are not independent tests. The two scenarios share 12 purchases and therefore do not provide separate confirmations of the strategy. Although the seven companies were identified before the measured purchases, the 25% rule was selected retrospectively. This experiment provides evidence worth considering, not assurance that the next dip will recover.
The spreadsheet appendix provides both starting scenarios, all 15 purchases, their execution dates, daily-price calculations and matched SPY comparisons. Historical stock prices and dividend adjustments come from archived Yahoo Finance data, supplemented for November 13 by Pubmarks and Yahoo Finance historical-price pages. SPY’s dividend-adjusted history, supplemented by Yahoo Finance chart data for May 22, 2023, supplies the benchmark. Returns use the providers’ reinvested-dividend adjusted-price convention; dividend reinvestment is not modeled as an additional external contribution



I’d like to turn this experiment into a contest: Who can design the best buy-the-dip rule? Try smaller firms, shallower or deeper buying thresholds, additional purchases after further declines, or rules that take some profits after a recovery. My 25% decline rule with no sales is the starting entry. We would need a common investment budget, a defined stock universe and rules for handling idle cash so the comparisons are fair. Historical results could identify promising contenders; tracking those rules going forward would show whether their success survives beyond the data used to design them. Reader suggestions welcome.