Pure Binary Moving Average Crossover Rules
This test removes all secondary tactical overlays—no RSI oversold bounces and no realized-volatility scaling. Every moving average formula is judged strictly on a binary daily close rule, and every return comes from the real ETF’s adjusted closes (not a synthetic reconstruction):
- Signal Close ≥ Moving Average: Hold 100% of the 3x long ETF (for example TQQQ) on the next trading day. The signal is computed on the underlying index or sector ETF (for example ^NDX for TQQQ), using its full history so the moving average is warmed up before the ETF launches.
- Signal Close < Moving Average: Rotate into your chosen defensive sleeve—100% Cash (earning the 13-week T-Bill yield), 100% TMF (real +3x 20+ Year Treasury ETF), or 100% of the exact -3x inverse ETF of the same index (for example SQQQ).
- Pick any pair: TQQQ/SQQQ, UPRO/SPXU, SPXL/SPXS, UDOW/SDOW, TNA/TZA, SOXL/SOXS, TECL/TECS, FAS/FAZ, LABU/LABD, YINN/YANG or TMF/TMV. Each backtest starts on the later launch date of the long and inverse ETFs (and TMF when selected).
Equal weight across all N days. Slowest to turn at peaks and bottoms, minimizing whipsaws in choppy markets.
Weights recent closes exponentially via k = 2/(N+1), reacting earlier than SMA to sudden market tops and bottoms.
Computes WMA(2·WMA(N/2) − WMA(N), √N) to strip out lag while staying smooth—testing raw speed vs. whipsaw drag.
Uses the Efficiency Ratio to speed up during clean directional trends and flatten out during sideways chop.
Subtracts single-EMA lag via 2·EMA(N) − EMA(EMA(N)).
Uses 3·EMA − 3·EMA(EMA) + EMA(EMA(EMA)) for ultra-low lag at trend inflection points.
De-lags price prior to smoothing by feeding Close + (Close − Close[(N−1)/2]) into an EMA.
Applies linearly decaying weights from N down to 1 across the lookback window.