How the models are built, tested, and run.
No black boxes. Every rule below is the actual rule the models trade against — not a simplified summary for marketing purposes.
1. Fixed rules, written before the fact
Every strategy is fully specified in code: entry trigger, exit trigger, the table that turns votes into an allocation, and the quarterly rebalance. Signals use only the previous session's close and are published before the next session opens. There is no discretionary override.
2. Ensemble voting, not single-indicator prediction
Each strategy combines 7-10 independent signals — trend, momentum, realized volatility, macro gates (VIX, credit spread, Fed Funds direction, CPI YoY) and asset-specific drivers (real rates, M2 growth, yield-curve shape, unemployment). The strategy is IN when enough signals agree and OUT when too few do, with a gap between the two thresholds so it does not flip on noise.
3. Allocation from the vote count
While a strategy is IN, how much of its bucket you hold depends on how many signals agree: full agreement → 100%, then 75%, then 50%; below the lowest tier → 0% (cash) even though the exit rule has not fired. The allocation is decided from the previous close and traded at the next open. Equity strategies use 7 votes (100% at 7, 75% at 6, 50% at 5); Treasuries the same table over 7 votes; gold uses 10 votes (100% at 9-10, 75% at 7-8).
4. Execution you can copy
ENTER: a buy stop at the signal day's high or close (whichever is higher) plus 0.15%, valid 3 sessions, then a buy at the close of the 3rd session if untouched. EXIT: sell at the next open. ALLOCATION CHANGE: trade the difference at the next open. No stop-loss orders and no intraday decisions. The backtest charges 0.10% slippage and 0.06% commission per trade and a daily financing cost on leveraged exposure.
5. What is real and what is simulated
You trade TQQQ (3× Nasdaq-100), UPRO (3× S&P 500), TMF (3× 20-year Treasury) and UGL (2× gold). TQQQ and TMF use the real ETF prices once the ETFs exist (2010 and 2009); before that, and for the whole history of UPRO and UGL, the series is simulated from the underlying index with daily-reset leverage, fees and financing. Simulated history is an estimate of what these instruments would have done, not a record of it. Cash is assumed to earn short-term interest rates.
6. Portfolio construction
Four equal 25% buckets, re-equalized on the first trading day of each quarter. Measured daily return correlations are low between the growth strategies and the Treasury and gold strategies (below about 0.1 in absolute value), which is why the blended portfolio has a much smaller drawdown than its leveraged parts. The statistics shown on each dashboard page are recomputed every trading day.