When Your Strategy Breaks...
StatsEdgeTrading
So you built a strategy. 15 years of backtesting. Two years of paper trading. Real capital deployed in January 2025. You’re right two-thirds of the time. Everything looks perfect—until it isn’t.
This episode picks up on a case study from a user named Andre, who did everything right. Then suddenly, performance tanked.
Key findings:
Not a sample size issue. With 200–370 trades per year and clean in-sample vs. out-of-sample logic, the system wasn’t underbuilt.
Win rate held. Profit factor didn’t. From 2.3 to 1.7 to 1.55—with costs tipping it toward breakeven. That’s not just variance. That’s signal decay.
Trade frequency spiked in 2025. One of the best debugging clues in system design. Rather than less edge, there might be too many low-quality signals now triggering.
Action Plan:
Start with your raw strategy, pre-optimization. Did base idea trade frequency also rise?
Recheck column importance. Some features may have lost predictive power. Others may now dominate.
Add new columns. This is where insight lives. What macro inputs should you have included?
Finally, build filters that drop the worst new trades and retrim to historical volume. This often restores edge.
Andre’s mistake wasn’t doing too little—it was doing so much that the system outgrew its scaffolding. But more trades mean more data, which makes this fixable.
If you’re running a strategy and seeing similar shifts—flattening returns, edge degradation, or signal volume explosions—it’s probably not random. It’s a structural shift you can debug.
For more case studies, algorithmic trading ideas, and swing setups:
👉 www.statsedgetrading.com