Hated Mag 7
◐ Regime-dependentEach month, buy the Magnificent 7 stock with the worst 3, 6 and 12-month returns. Rank AAPL, AMZN, GOOGL, META, MSFT, NVDA and TSLA on each window (1 = best), average the three ranks, and put everything into the worst. Rerun at every month-end.
Right now
Ranking at the latest close
| Stock | 3M | 6M | 12M | Avg rank |
|---|---|---|---|---|
| MSFT | +38.7% #1 | +41.6% #1 | +2.0% #5 | 2.33 |
| AAPL | +20.3% #3 | +35.1% #3 | +35.7% #2 | 2.67 |
| META | +36.7% #2 | +37.5% #2 | -0.8% #6 | 3.33 |
| NVDA | +17.0% #4 | +31.7% #4 | +27.5% #3 | 3.67 |
| GOOGL | +2.0% #6 | +22.6% #5 | +39.5% #1 | 4.00 |
| AMZN | +7.3% #5 | +20.3% #6 | +13.4% #4 | 5.00 |
| TSLA | -2.0% #7 | +0.0% #7 | -16.0% #7 | 7.00 |
Rank 1 = best return. The bottom row is the most hated, which is what the rule buys. Ties go to the worse 12-month return.
Get the monthly signal
One email after the last trading day's close each month: hold, or switch and to what. Confirm by email; unsubscribe in one click.
Live record
Written by the bot as it trades, starting on launch day. Rows are never edited or deleted. It follows the published timing: signal at the month-end close, trade at the next day's close.
| Date | Action | Sold | Bought |
|---|---|---|---|
| Fri, 25 Sept 2026 | Start | – | META @ 751.66 |
Backtest lab
Loading prices…
Benchmarks start on the same day. Equal-weight rebalances on the strategy's own trade dates; SPY is bought once. Costs apply to all of them, charged on turnover (10 bps = one full round trip).
Drawdown from peak
What it held
Monthly holdings log
Does the start month matter?
CAGR minus equal-weight Mag 7's CAGR, in percentage points a year, for every possible first signal month, using the lab's current settings. The selected start is outlined.
Does the viral claim check out?
The post claimed returns "since 5/23". First signal 2023-05-31, measured to the 2026-08-31 rebalance.
Trading at the signal close reproduces the post almost exactly. Trading at the next close, which is what a bot running after the close can actually do, happens to do better over this window.
Verdict
The numbers are real; the edge is regime-dependent. Since May 2023 the rule beat equal-weight Mag 7, the "most loved" control and SPY for every start month we tried, and 10 bps of costs barely dent it (the picks are sticky: 20 switches in 40 months).
Over 2013–2026 it's ordinary. Its CAGR beats equal-weight, but with about 40% more volatility and a lower Sharpe ratio. It doesn't beat the "most loved" control either: level with it when trading at the next close, half its total return when trading at the signal close. Before 2023 it beat equal-weight in only 4 of 10 calendar years.
Caveats
- Hindsight basket. "Mag 7" was coined in 2023. A rule that only ever buys stocks we already know became $1T+ companies looks better in the past than it can in the future. Its weak spot is a member that falls out of the club for good.
- Small sample. 40 monthly picks since May 2023. One month, or trading one day later, moves the total by 100+ percentage points.
- One stock at a time. Volatility runs 37–39% a year and drawdowns of 30–45% are normal.