Backtest Validation

Test design

The backtest simulates the QQQ LEAPS strategy from January 2021 through August 2026. A $10,000 reference account is assumed, with per-position size capped at one third of the account, at most three positions open at once, and a 5% cash reserve. Entry and exit rules are fixed before each walk-forward window, and the ML model is retrained only on data that existed before the period it trades. Commissions and slippage are charged per fill at standard retail option rates, and margin treatment follows standard Reg-T assumptions.

Results

Hypothetical or simulated performance results have certain inherent limitations. Unlike an actual performance record, simulated results do not represent actual trading and may under- or over-compensate for the impact of certain market factors. Past performance, real or simulated, does not guarantee future results.

The reality check we publish next to it

A model backtest says the worst peak-to-trough drawdown was 14.5%. A separate 15-month test of the same rules priced with real market quotes drew down 30.4%. Those are different windows and different pricing, and both numbers are published, side by side, on the landing page and in every issue that mentions performance. If you are comparing us to someone who shows only the better number, ask where the other one is.

Comparisons and stress tests

The same window is evaluated against buy-and-hold QQQ, a static LEAPS position without the signal gates, and the strategy with the ML layer removed. The rules were also stress-tested across 21 recombined sub-windows (including 2022's selloff) and held up in 18. Every fill, win or loss, is in the public ledger, including the days the system chose to do nothing.

Limitations

• Model-priced fills: backtested entries and exits are priced by a math model, not by live market quotes. Real fills will differ, sometimes meaningfully.
• Survivorship of the test window: January 2021 to August 2026 contains specific regimes. A different window would produce different numbers.
• Liquidity assumptions: the model assumes orders fill at modeled prices. Fast markets and wide spreads are only approximated.
• Parameter selection: every rule was chosen at some point by a human looking at data. Walk-forward testing reduces, but cannot eliminate, this bias.
• A backtest cannot see the future: no amount of validation makes a simulated result a promise.