How the Synthetic TQQQ / SQQQ Tactical Strategy Works
This strategy uses daily closes of the real Nasdaq-100 (^NDX) back to October 1985 to simulate a low-turnover TQQQ / SQQQ tactical system (~9–10 trades per year on average), using Synthetic TMF (20+ Year Treasuries modeled from ^TYX) as a volatility shock absorber rather than chasing every minor dip.
- 150-Day SMA Trend Core (±3.0% Hysteresis): Holds TQQQ when ^NDX closes more than 3.0% above its 150-day Simple Moving Average, and flips to SQQQ only when ^NDX falls more than 3.0% below it. That 6% deadband eliminates noisy whipsaws around the moving average.
- Oversold RSI(10) Bounce (< 30) with 2.0% Stop-Loss: Even in a bear regime below the SMA, if 10-day Wilder RSI drops below 30 (capitulation selling), the strategy temporarily rotates back into TQQQ to catch sharp V-bottom rebounds (promoting to Bull Trend if ^NDX reclaims the 150d SMA). A tight 2.0% ^NDX trailing entry stop-loss exits back to SQQQ if the bounce fails.
- 25% Volatility Cap & 25% Rebalance Deadband: Measures 10-day annualized realized volatility of ^NDX. When vol exceeds 25%, equity leverage is scaled down (
min(100%, 25% / vol), floored at 15%) and the remainder is parked in TMF. To keep turnover low, the weight only rebalances when target allocation shifts by at least ±25 percentage points. - 25% Bubble Trim: If ^NDX stretches more than 25% above its 150-day SMA during a parabolic bull run, TQQQ exposure is capped at 50% (with the other 50% in TMF) to cushion sudden mean-reversion air pockets.
- 50% Extreme-Vol Crisis Shield: When 10-day realized volatility spikes above 50% (such as 1987, 2008, or March 2020), the strategy steps aside from both TQQQ and SQQQ into 100% TMF until panic subsides, avoiding violent short-squeeze whipsaws.
- Realistic Synthetic Drag & ^IRX Financing: Daily +3x and -3x returns include the 0.88% ETF expense ratio plus dynamic historical swap financing modeled from actual 13-week T-Bill (^IRX) rates and an institutional 0.45% swap spread.
- Short gate (200-day SMA): The strategy only shorts via SQQQ when ^NDX is below both its trend SMA band and its 200-day SMA. When ^NDX is below the trend band but still above its 200-day SMA (a sharp dip inside a long uptrend, like July 1996 or October 1998), it holds 100% TMF instead of shorting. This avoids whipsaw losses from V-shaped dips: in our 1985–present tests it added about 3–5 points of CAGR, almost all before 2010, with equal or slightly shallower drawdowns.
- Crisis panic buy (RSI < 35 for 2 days): While the crisis shield has the portfolio in TMF, if 10-day RSI on the index stays below 35 for two days in a row (a capitulation low), the strategy moves half the portfolio back into the 3x long ETF (50% long / 50% TMF) until the shield ends. In backtests this adds 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 (Historical US Rates): Off by default (assuming a tax-advantaged Roth/IRA). When enabled, tracks FIFO tax lots and settles annual taxes using historical 1985–present US federal top short-term (28%–50% + 3.8% NIIT since 2013) and long-term (15%–28% + 3.8% NIIT) capital gains rates with loss carryforward.
Why These Settings? Robustness Notes
We stress-tested these rules on 1985–present ^NDX using a simplified 3x simulation (TMF sleeve treated as cash, rough financing costs), changing one thing at a time. Absolute returns differ from the backtester above, so treat the direction of each result as the takeaway. Backtests are not predictions.
- RSI threshold: 28–31 is a stable plateau (about 43–48% CAGR in the test). Below 27 it collapses (17–34% CAGR at 24–27) because too few bounces fire; above 32 it fades to about 35–41%. If you adjust it, err higher rather than lower.
- 2.0% bounce stop: a local optimum. Tighter stops (0.5–1.5%) get chopped out by normal rebound noise; wider stops (3–10%) push max drawdown to −90% or worse.
- No “structural bear” filter: switching the bounce off during structural bears (falling 200d SMA, death cross, or >25% below the 1-year high) cut CAGR from about 48% to 20–47% and deepened drawdowns. Bear-market bounces win less often (about 27%) but the winners are large and the 2.0% stop caps the losers.
- Single 150d SMA ±3%: beat 100/125/175/200-day alternatives and 15 multi-SMA ensembles (majority vote, unanimous, averaged), which lag turning points.
- No VIX filter: requiring a VIX spike or a VIX roll-over before taking the bounce cut CAGR to roughly 6–35%. RSI already captures the capitulation.
- Vol cap, crisis shield, bubble trim: results moved by only about 1 point of CAGR across 20–40% vol cap, 50–60% crisis threshold and 25–40% bubble trim. A 20% vol cap trades roughly 1 point of CAGR for a shallower drawdown (about −71% vs −79%).
Known Weakness: Whipsaws in Fast V-Shaped Dips (the 1990s Boom)
Zoom to 1995–1999 and the strategy visibly lags buy-and-hold (roughly 67% a year versus 168% with the auto-tuned settings). The cause is a trade-off built into any 150-day SMA rule rather than a bug, and we tested the obvious fixes on the full 1985–present history using a replica of this engine.
- What goes wrong: The 1990s dips (July 1996, April 1997, late 1997, August–October 1998) were sharp and recovered within weeks. Each time ^NDX closed more than 3% below its 150-day SMA the strategy flipped to SQQQ and was then run over by the rebound. In 1995–1999 it switched regime about 32 times, and the roughly 59 days spent in SQQQ lost about 57% while buy-and-hold gained about 90% over the same days.
- Bubble trim adds to it: In a parabolic run the index sits far above its SMA for long stretches. A tight 15% bubble trim caps TQQQ at 50% there, which cost roughly 20 points of 1990s CAGR in the test (about 87% without the trim). Over the full history the effect is small (about 1 point).
- Shorting only below the 200-day SMA works best: Holding TMF instead of SQQQ unless ^NDX is also below its 200-day SMA added about 3–5 points of full-history CAGR with equal or slightly shallower drawdowns (for example 49.7% to 54.8% on the tuned settings). It lifted 1995–99 from about 67% to 109% and the 2000–02 crash from −13% to −6%, left 2008 and 2010-onward unchanged, and gave up about 5 points over the last 10 years.
- Waiting for several closes is not worth it: Requiring 3–5 closes below the band before flipping helped the 1990s and the crashes but cut the 2010-onward CAGR from about 67% to 53–57%, because it enters real bear markets late.
- Never shorting is worse: Replacing SQQQ with TMF in every bear dropped full-history CAGR from about 50% to 43%. SQQQ is what makes money in 2000–02, so the lesson is to short less, not to stop shorting.
- Why the page keeps the simple rule: The 200-day gate changes nothing on 2010-onward data, so it would not help the real-ETF pages, and all of these results are in-sample. This page now includes the 200-day short gate described above; the real-ETF pages do not, because it changes nothing on 2010-onward data.