The rules
Buy at the next minute's open after an hourly close reaches the upper band: the lower of the highest of the last 20 session closes and their 20-session average plus 2.8 times their mean daily change. Hold it with a trailing stop at the lower band, the higher of the lowest of the last 40 closes and their 40-session average less 2.8 mean changes, which rises with each session and never falls; an hourly close below it sells. The averages are exponential. Long only, 8.2% of the account per stock.
- Universe
- The twelve largest holdings of Global X's AI and technology fund (AIQ) on November 5, 2026: MSFT, PLTR, AMZN, TSLA, AAPL, ORCL, NFLX, GOOGL, NVDA, SAP, TSM, CSCO
- Bars
- hourly
- Position size
- 8.2% of equity per stock at entry, never rebalanced
- Rebalance
- Checks every hourly close; the stop moves once a session
Results
| Measure | Orbital test | SPY, same days |
|---|---|---|
| Window | Oct 30, 2023 to Oct 30, 2026 | Oct 30, 2023 to Oct 30, 2026 |
| Total return | +77.30% | +88.60% |
| CAGR | +21.03% | +23.54% |
| Sharpe | 0.98 | 1.22 |
| Discounted Sharpe | −0.70 | — |
| Max drawdown | −15.5% | −18.7% |
| Win rate | 40.1% | — |
| Profit factor | 1.98 | — |
| Trades | 137 | — |
Hypothetical: a backtest with modeled costs, not traded with real money.
Hypothetical growth of $10,000
I13 AI leaders, Keltner-Donchian trend · +77.3% vs SPY +88.6%
Rebuild it in Orbital
Paste this into Orbital Chat. The Strategy Developer builds the workflow and asks before it runs any backtest.
Trade MSFT, PLTR, AMZN, TSLA, AAPL, ORCL, NFLX, GOOGL, NVDA, SAP, TSM and CSCO on 1-hour bars, built as a program that keeps session closes. Upper band: the lower of the 20-session high close and the 20-session EMA plus 2.8 times the mean absolute daily change. Lower band: the higher of the 40-session low close and the 40-session EMA less 2.8 mean changes. Buy on an hourly close at or above the upper band and trail a stop at the lower band that never falls; sell on an hourly close below it. 8.2% of equity per stock, long only. Backtest from October 30, 2023 to October 30, 2026 with $100,000.
Is it real?
- Strategies tested together
- 15
- Beta to SPY
- 0.69
- Alpha vs SPY
- +3.7% a year
- Costs paid
- $1,680, 2% of the gain before costs
- Data
- 1h decisions on 1-minute bars, 754 sessions
- Fills
- Next open at market, sized at the fill
Read this before you trust it
This basket was picked in November 2026, with hindsight: a fund's biggest holdings today are the companies that grew, which tilts any backtest of them toward beating SPY. It runs beside the same rules on the October 2023 Nasdaq-100, which has no hindsight, to show how much of a result the basket carries. Picked on November 8, 2026 by a rule written before any run: published evidence, yearly costs at most a quarter of the published gross return, buildable in Orbital, and the source's own parameters. Every rule is published whatever it shows. Orbital's engine cannot rank entries for a program that decides more often than daily, so instead of the source's position cap each stock gets an equal share and every signal is taken. Orbital's minute data starts in September 2023, so the session records start on October 2, 2023: early in the test a record is a short one. The selection expected it to trail SPY's return with a smaller drawdown, as the paper's own tradable version did.