Momentum-ranked ten-day trend hold, big tech 7

The rules

At each day's close it checks each of seven big tech stocks it does not hold, buys at the next open each one that closed above its 20-day average without unusually wide daily swings, and sells it at the open ten trading sessions later. A stock qualifies when today's close is above the average of its last 20 closes, today's included, and its 14-day average true range is at most 10% of today's close. A day's true range is the largest of its high minus its low and the distances from the previous close to its high and to its low; the 14-day figure is the plain average of the last 14. Each buy is 20% of what the account was worth at the previous session's close (the starting cash on the first day), in whole shares at today's closing price, and fills at the next session's open. There is no stop loss or profit target, and a stock sold at the open can qualify again at that day's close. With no margin, it held at most five stocks at once. When more qualify than the cash can buy, they are bought in order of a score, highest first: the 20-day return, today's close against the close 20 sessions earlier, divided by the average true range as a share of today's close. An order the remaining cash cannot cover in full is skipped, not shrunk, and the next in line is tried; nothing held is sold to make room. At each open the sales come before the buys, so the cash a sale frees can pay for a buy at the same open. The score only sets the order, so a stock down over 20 days can still be bought if it closes above its 20-day average. A stock is also skipped for the day if any of its last 33 daily bars fails a data check (a missing price or volume, a low at or below zero, or an open or close outside that day's high-low range) or its average volume over the 20 sessions before today is zero. All counts use the stock's own trading sessions, and a half day counts as one. The program also carries a small forecasting model, switched off in this version; all it still requires is 20 past examples to learn from, which the 64 sessions of history before the test provide by the first close. Buying can start at the first close, January 2, 2020, with the first fills at the next open, January 3; positions still open at the end are sold at the last close, September 4, 2026, and buys decided at that close never fill.

Universe

Seven big tech stocks: AAPL, MSFT, NVDA, AMZN, GOOGL, AMD and TSLA

Bars

daily

Position size

20% of the account's value at the previous session's close per stock, in whole shares; with no margin, five positions take about the whole account, and it never held more than five

Rebalance

Every session: decides at the close and buys at the next open; each stock is sold at the open ten sessions after its buy

Results

Hypothetical: a backtest with modeled costs, not traded with real money.

Hypothetical growth of $10,000

Rebuild it in Orbital

Paste this into Orbital Chat. The Strategy Developer builds the workflow and asks before it runs any backtest.

Trade AAPL, MSFT, NVDA, AMZN, GOOGL, AMD and TSLA on daily bars during regular hours. Build it as a program that uses only daily data, decides at each session's close and declares a 64-bar warm-up, so it can trade from the first session. At each close, for each stock with no shares held and no order pending, compute the average of its last 20 closes, today's included; its 14-day average true range, the plain average of its last 14 true ranges, where a day's true range is the largest of its high minus its low, the distance from the previous close to its high and the distance from the previous close to its low; and its 20-day return, today's close divided by the close 20 bars back, minus 1. Buy with 20% of equity when today's close is above the 20-day average and the average true range is at most 10% of today's close. Skip the stock that day if any of its last 33 bars has a missing price or volume, a low at or below zero, or an open or close outside that bar's high and low, or if its average volume over the 20 bars before today is zero. Give each buy a score equal to the 20-day return divided by the average true range as a fraction of today's close; a negative score still buys. Set a 10-bar time stop on each buy, so each sale fills at the open ten sessions after the buy, with no stop loss and no profit target. Orders fill at the next bar's open. Rank each day's buys by score, highest first, with at most 7 new positions a day, at most 7 held at once and no minimum score, so when cash runs short the highest scores are bought first and nothing held is sold to make room. Long only, on a cash account with no margin. Backtest from January 1, 2020 to September 4, 2026 with $100,000.

Is it real?

Versions tried

96 across the 11 "Stocks" research strategies

Discounted Sharpe

−7.24: the raw 1.10 less the 8.34 the best of 96 tries would reach by luck

Chance it beats luck

less than 0.1% that its true Sharpe tops 8.34

Beta to SPY

0.85

Alpha vs SPY

+18.0% a year

Data

Daily bars, 1,678 sessions

Costs

Alpaca's $0 commission, 0.10% slippage a side

Read this before you trust it

96 versions were tried across the 11 "Stocks" research strategies, and their Sharpes ran from −12.76 to 1.53. The best of 96 tries with no real edge would be expected to reach about 8.34 by luck alone; this run's 1.10 falls 7.24 short of it, and the chance that its true Sharpe is above it is less than 0.1%. Many versions ran over windows of six months to a year, whose Sharpes swing far more than a multi-year run's; that spread raises the bar, so the discount errs on the strict side. The seven stocks were picked in September 2026, with hindsight of how large tech stocks had done since 2020, and the strategy held four or five of them on about three sessions in four, so its result rests largely on how these stocks did. The run assumed slippage of 0.10% a side, twice Orbital's default of 0.05%; a backtest from Chat uses the default, so a rebuild will likely come out somewhat higher. Its deepest fall from a high, about 38%, ran from November 19, 2021 to November 9, 2022, and the high was not regained until June 14, 2023.