Leveraged S&P 500 ETF Strategies: What Worked on UPRO, SPXL and SSO, and What Did Not

Four tests, each on every fund. Every dollar figure is the profit on a steady $10,000 in each trade, after one cent of slippage on every market and stop order.

Per $10,000 in each trade on UPRO, June 2009 to October 2026, profits not reinvested. The last column runs the same rule on SPY since 1993.

Sharpe: the yearly profit against the day-to-day swings — higher is smoother. Worst fall: the largest drop from a high, in dollars. Sized so its worst fall matched holding’s, the night-only dip buy would have made about $7,100 a year — but that takes 5.3 times the money in every trade. The 200-day rule held day and night gives half the fall of holding for half the profit; at the same fall it makes about $3,900, about what holding makes.

RSI(3) is the Relative Strength Index of the last 3 daily closes, a 0–100 gauge of recent ups against downs. Under 20 means the fund has dropped hard for a few days. The rule buys that drop and sells when it fades.

1. 1

   At the 4:00 pm close

   If the fund’s RSI(3) is under 20, buy at the close. A market-on-close order fills at the closing auction price.

2. 2

   Night-only version: the next morning at 9:30

   Sell at the open with a market-on-open order. On UPRO: $1,352 a year per $10,000, a worst fall of $2,183, up in 17 of 18 years.

3. 3

   Hold version: the first close with RSI(3) at 20 or more

   Sell at that close, usually 2–3 days later. On UPRO: $2,083 a year per $10,000, a worst fall of $5,024, up in 16 of 18 years, in the market about 1 day in 5.

An overnight hold is not a day trade, so the limit on day trades in small margin accounts does not apply. The test reads the signal at the close and trades at that close; it charges a cent of slippage each way, which market-on-close and market-on-open orders mostly avoid.

Profit per $10,000 in each calendar year (2009 from June 25, 2026 to October 13). In 2022, when holding lost $5,783, both RSI(3) rules made money.

Years up: holding 16 of 18, the RSI(3) hold 16 of 18, the RSI(3) night hold 17 of 18, the overnight hold above the 200-day 12 of 18.

1-minute prices, January 2016 to October 2026, per $10,000 per trade. For scale: random long trades held 2 hours reached a Sharpe ratio of only 0.31 on UPRO at the 95th percentile.

The top two made money in both halves (2016–21 and 2021–26) on all four funds. They are day trades: in a margin account under $25,000 that means 3 in 5 business days; a cash account can make one round trip a day with settled cash.

Search 1,764 rules and some will look great by chance. These checks separate a pattern from a lucky pick.

• !Mostly a bull market: the real funds’ dollar figures come from 2009 on. The built funds since 1993 include the 2000–02 and 2008 bear markets, and the dip and overnight rules passed there too — but holding a built 3× fund since 1993 fell as far as $21,695 on a position kept at $10,000.
• !Leverage magnifies bad years: holding UPRO lost $5,783 per $10,000 in 2022 and $1,587 in 2018.
• !Costs could be higher: the test charges one cent of slippage each way; thin trading before the open or a wide spread on a crash day would cost more. The overnight rules trade most nights, so every cent counts.
• !Day-trade limits: in a margin account under $25,000, only 3 day trades are allowed in 5 business days; a cash account can make one round trip a day with settled cash.
• !The two day-trading rules that held up were designed on Nasdaq futures and QQQ over the same decade, so the S&P funds are a related market, not an independent test.
• !Sizing up to holding’s risk is only on paper: matching holding’s worst fall with the night-only rule means 5.3 times the money in every trade, on a fund that is already 3×.

• Prices: real daily prices with dividends added back for SPY (from 1993), SSO (2006), SPXL (2008) and UPRO (2009) to October 13, 2026; 1-minute prices for all four from January 2016 (2,711 trading days each), set back to the prices they traded at before each split, so a cent is the cent it was.
• Built funds: SPY’s daily move × 2 or × 3, less a 0.9% yearly fee and the cost of the borrowed part (the Treasury-bill rate + 0.5%), back to 1993. Where they overlap the real funds they track them to within 0.2% a year.
• Money: a constant $10,000 in the fund for each trade (profits not reinvested), no commission, one cent of slippage on every market and stop order. $10,000 of a 3× fund is about $30,000 of S&P 500 exposure.
• Fills: a daily signal at the 4:00 pm close trades at that close or at the next 9:30 open, so a next-morning entry never gets the overnight move for free.
• “A year” is the average yearly profit; the Sharpe ratio is that profit against the swings of the daily dollars (higher is smoother); “worst fall” is the largest drop from a high; “years up” counts calendar years with a profit, the first and last partial.
• Checks: two out-of-sample tests (built 3× 1993–2008 → UPRO 2009 on; UPRO 2009–2017 → 2018 on), random entries made as often and held as long, and a profit in both halves of each history (the 1-minute tests split at May 31, 2021).
• Not included: the Fear & Greed index was not part of the indicator search.