Is machine learning actually useful for retail trading bots?

I wouldn’t claim ML has improved my live profits without verified results. I’d start with gradient boosting using features like volatility, ATR, spread, trend strength, and session timing. I’d use walk-forward validation, out-of-sample testing, and realistic transaction costs to limit overfitting. The real test is whether the model stays profitable in live trading across changing market conditions, not just whether its backtest looks impressive.