The code worked, but my equity curve didn’t behave like the backtest at all. What helped me was separating “strategy logic risk” from “portfolio risk.” For example, instead of fixed lots, I moved to volatility-based sizing and capped risk per trade across all systems combined, not per strategy. I also started stress-testing strategies with random delays and higher spreads to see how fragile they were. There’s a good breakdown of different quant approaches and their weaknesses here: https://forextester.com/blog/quant-trading-strategies/ — not as a holy grail, but it helped me rethink how fragile some models are. Since then, my drawdowns became boring, which is actually a good thing.