Mechanism
Strategies harvest specific conditions: trend strength, volatility level, correlation structure, liquidity.
When the regime changes, signal-to-noise changes. Your thresholds stop matching reality. Costs become larger relative to expected edge.
The strategy didn’t become ‘bad’. The environment changed.
Testing reality: backtests hide execution, spreads, and regime clustering. If your edge survives only in clean data, it’s a data story, not a trading edge.
- Test with variable spreads and slippage by session.
- Stress the tails: worst 1% days matter more than average days.
- Prefer rules you can execute consistently over rules that optimize the past.
How it kills accounts
Strategy works → confidence rises → regime shifts → losses cluster → trader tweaks → overfits → abandons at bottom → repeats the cycle.
How it kills accounts:
- Model works on clean history.
- Regime changes and execution friction increases.
- Performance decays slowly, so you rationalize.
- You optimize parameters instead of reducing risk.
- Drawdown becomes the teacher.
Rule that survives
Define the regime your strategy needs.
Track regime signals (volatility, trend, correlation).
When the regime is absent, reduce frequency and size.
Rule that survives:
- Stress test tails and execution, not averages.
- Prefer robust plateaus over optimized peaks.
- When performance decays, reduce risk before “fixing” the model.
Example archetype
A mean reversion system thrives in low vol and tight ranges. A trend regime begins; it gets steamrolled. You keep forcing it because ‘it always worked’. That’s not discipline; it’s denial.
Tell: if small parameter tweaks flip your results, your model is fragile.
Deep dive
Make regime explicit
Write down the conditions your system needs. If you can’t, you are trading a story, not a model.
Related: Data is not neutral.
Glossary: regime, volatility.
Field checklist
- Stress-test the tails. The worst days define survival.
- Use variable spreads and slippage in testing.
- Prefer robust plateaus over optimized peaks.
- Look at drawdown shape, not only profit.
- If a tiny parameter change breaks the model, the model is fragile.