TV Optimizer The overfitting trap
Test enough combinations on one period and something will look spectacular by pure chance. That configuration did not find an edge — it memorized history. The moment you trade it live, the magic disappears.
This is the single most common way optimization goes wrong, and it is why an optimizer without validation tooling is dangerous.
How walk-forward works in TV Optimizer PRO
After a run, the walk-forward mode takes your top configurations and re-evaluates each one across multiple time windows — for example, each of the last five years separately. You get a config × window matrix of real backtests.
You choose how to crown the winner: best combined Net Profit + Drawdown + Profit Factor, best average across windows, or best worst-window (the most conservative). A configuration that wins in 2021 and dies in 2022-2025 gets exposed immediately.
Robust zones beat sharp peaks
The plateau-robustness metric scores each configuration by how well its neighbors perform. If length 21 is great but 19 and 23 are disasters, that peak is probably noise. If the whole 18-26 zone performs, the edge is more likely real. Prefer plateaus; trade the middle of them.