The approach
The first question isn't whether your model works.
It's whether you're asking it the right question.
Three questions determine whether a strategy has real edge. Each one is answered by the same underlying process - evolutionary search across combinatorial spaces that grid search can't cover.
Question one
What should you be optimizing?
The same data can answer very different questions. Most people pick a target and start building. But if the target is wrong, everything downstream inherits that mistake - features, validation, sizing, all of it.
On a recent engagement, we tested five target definitions against the same dataset. Four produced no out-of-sample signal. The one that survived wasn't the one the client expected.
How the search answers this
The GA tests multiple target definitions simultaneously - Sharpe, sizing efficiency, drawdown, trade frequency, book width - alone and in multi-objective combinations. Different targets explore fundamentally different solution spaces. The search discovers which questions the data can actually answer, not which one you assumed.
Question two
What's signal and what's noise?
Most of what you're watching is the same three or four economic forces measured twenty different ways. Your model treats each one like independent information. It isn't.
The result is a model that's overconfident. It sizes positions based on a conviction level that doesn't exist - because twenty inputs that agree aren't twenty confirmations. They're one signal echoed twenty times.
How the search answers this
The GA's fitness function penalizes correlated features. Populations evolve toward decorrelated input sets - hundreds of candidates compress to the handful that survive selection pressure. Each surviving feature represents an independent economic concept, not the same signal measured differently.
Question three
When does it break?
If the features your model relies on stopped working today, how long would it take you to notice? Not whether it would happen - it will. How long between the signal dying and you realizing it.
Most traders find out after months of unexplained underperformance. By then the drawdown is deep and the conviction is gone.
How the search answers this
The process re-runs on expanding windows. When what mattered stops mattering, the search surface shifts - feature importance changes, fitness degrades, the surviving population looks different. You see the regime change in the search dynamics before you see it in the P&L.
Same process. Three questions. If the edge is real, you'll know.
Every engagement starts here. If we can't find signal, we say so. See services or get in touch.