Defensive Strategy
ML Defensive shares the same LightGBM sector-ranking stack as ML Aggressive, but it adds a continuous regime overlay that can scale total invested capital from roughly 15% to 100% based on market conditions. The bot still selects the top three ETFs from the 14-name universe, but when conditions deteriorate it deliberately leaves more of the portfolio in cash equivalents and parks residual capital in SGOV. The result is a lower-beta, capital-preservation version of the ML engine rather than a separate stock-picking model.
What it does
It ranks the same 14 ETFs as the aggressive bot, but then uses a regime model to decide how much capital should actually be risked before sweeping the rest into SGOV.
Best for
Investors who want the ML rotation engine with more explicit downside control and a built-in risk-off posture.
What to expect
Lower market exposure during stressed regimes, frequent periods with a meaningful SGOV sleeve, and smoother behavior than the aggressive bot in weak tapes.
Representative basket only. The live bot dynamically holds the current top 3 model picks and can shift a large share of capital into SGOV when the regime score turns defensive.
Compare with
Normalised to $10,000 start. Data from live Alpaca paper account — reflects real orders, real fills, and real risk management decisions.
Simulated using a multi-timeframe momentum proxy and inverse-volatility sizing — approximates the ML signal engine. Run scripts/generate_ml_backtest.py for exact results.
ETF Momentum
A rules-based sector rotation bot that applies an A5 momentum filter, ranks qualifying ETFs by relative strength vs SPY, and parks weak-signal capital in SGOV.
ML Aggressive
A machine-learning sector rotation bot that ranks a 14-ETF universe, holds the top 3 names, and stays effectively fully invested without regime scaling.
Bangkok
A hybrid bot combining weekly ETF momentum rotation with a Kestrel swing-trade stock sleeve, dynamically funded from confirmed paper gains.
Ready to put this strategy to work?
Open a paper position and track it against the market — no real money at risk.
Live chart: real Alpaca paper account data. Simulation chart: historical price data from Yahoo Finance with representative fixed weights. Past performance does not guarantee future results.