How the Tactical 3x ETF Backtester Works
Pick any of 10 real 3x ETF pairs. Every return comes from the actual ETFs’ adjusted closes (so expense ratios, swap financing and tracking error are already in the prices), while the signals are computed on the underlying index or sector ETF using its own longer history, warmed up before the ETFs launch. The same low-turnover Flagship rule set runs on every pair, and a signal at the close of day t governs the return on day t+1 with no look-ahead. Auto-Tune runs when you load a pair and whenever you change its goal or window; changing a slider by hand switches to custom mode until you re-tune or reset.
The Pairs
The Rules
- SMA trend core (±3% hysteresis): Holds the 3x long ETF when the signal series closes more than 3% above its SMA, and flips to the exact inverse ETF only when it falls more than 3% below. The 6% deadband removes noisy whipsaws.
- Oversold RSI(10) bounce with a 2% stop: Inside a bear regime, if 10-day Wilder RSI on the signal series drops below the threshold (capitulation selling), the strategy rotates back into the 3x long ETF to catch the snap-back. It returns to the bull regime if the signal reclaims its SMA, and exits to the inverse ETF if the bounce fails and the signal drops 2% from entry.
- Volatility cap with a 25-point rebalance deadband: When 10-day realized volatility exceeds the cap, exposure scales to
min(100%, cap / vol)(floored at 15%) and the remainder sits in TMF. Weights only rebalance when the target shifts by 25+ points. - Bubble trim and crisis shield: The long ETF is capped at 50% (the rest in TMF) when the signal stretches far above its SMA, and the whole portfolio moves to TMF when 10-day volatility spikes past the crisis level.
- Crisis panic buy (RSI < 35 for 2 days): While the crisis shield has the portfolio in TMF, if 10-day RSI on the signal series stays below 35 for two days in a row (a capitulation low), half the portfolio moves back into the 3x long ETF (50% long / 50% TMF) until the shield ends. In Nasdaq-100 backtests this added a few points of CAGR, mostly from the V-shaped rebounds of 2000–02, 2008 and 2020, without deepening drawdowns; the edge is modest, in-sample, and depends on a handful of crisis episodes.
- Optional taxable account: Off by default (Roth / IRA). When enabled, tracks FIFO tax lots and settles annual taxes with historical US federal short- and long-term rates plus NIIT and loss carryforward.
How Auto-Tune Works
- Search: It scores several thousand combinations of SMA length, RSI level, volatility cap, bubble trim and crisis level for the selected pair and applies the best one.
- Goals: Maximum Raw CAGR chases the highest return. Drawdown Shield looks for the shallowest drawdown among settings that still beat buy-and-hold.
- Windows: All-Weather blends the full history (40%) with the last 10 years (60%) and penalizes settings that lose to buy-and-hold recently. The others score only the full history, the last 10 years, or your chart zoom.
- Tune All Pairs: runs the search on every pair and lists each winner next to its buy-and-hold result, so you can compare pairs at a glance.
What to Expect
- Results differ a lot by pair. Pairs with deep, fast V-shaped dips and strong trends (semiconductors, technology) tune very differently from slower or choppier ones (financials, China, biotech). Treat the Tune All Pairs table as a comparison of what each history allowed, not a ranking of future performance.
- The tuned settings are a peak, not a plateau. In our Nasdaq-100 sensitivity tests on 2010-onward data, a 150-day SMA returned about 62% a year while 100, 125, 175, 200 and 220-day SMAs returned about 45–54%. Every trend length still beat buy-and-hold, but expect results nearer the middle of that range than the headline number.
- The RSI bounce is the biggest lever. Turning it off cut the Nasdaq-100 backtest from about 62% to about 11% a year, while moving the threshold by five points cost 15–35 points. Leave this setting near its tuned value.
- Vol cap, crisis shield and bubble trim matter less. They trade a few points of CAGR for drawdown protection and moved results by roughly 1–5 points across reasonable settings.
Things to know: TMF is the hedge sleeve for every pair, so each backtest starts no earlier than TMF’s April 2009 launch. Settings are tuned in-sample on a short live history, so treat them as the best this history allowed rather than a forecast. Sector pairs use the sector ETF (XLK, XLF, XBI, FXI) as the signal series, not the exact index the leveraged fund tracks.