What Test Lab is
Test Lab is where a system is researched before it is trusted. You run it on historical candles, stress it, benchmark it, and get an honest read on whether it is ready — all before a dollar of real capital is at stake. The goal is not a pretty curve; it is to surface weakness early.
The tools
- •Backtest — run the system on real historical candles and see how it would have behaved.
- •Walk-forward — the honesty test. The system is validated on data it was not tuned on, in rolling windows, so a result that only works on one lucky period is exposed.
- •Buy-and-hold benchmark — compare the system against simply holding the asset. If passive holding beats it, the complexity is not earning its keep.
- •The proof ladder — five rungs that stand between a system and capital: the blueprint (intent written down, including why NOT to launch it), a backtest, validation on live data (paper), a passed Risk Gate, and downside protection the engine actually executes. All five, or the system is not proved.
- •Deploy from a saved run — when a run looks solid, promote it straight into a live system with the same parameters.
In-product example
Backtest a grid over the last quarter, then walk-forward it across three rolling windows. If it prints profit in-sample but falls apart out-of-sample, that gap is the system telling you it is curve-fit — better to learn that here than with money.
The pitfall
The classic trap is overfitting: tuning parameters until the backtest looks perfect on one window. A curve that only works on the data it was fitted to will not survive live. Trust walk-forward over in-sample, respect the buy-and-hold benchmark, and treat a great single backtest with suspicion, not excitement.
Next
Capital comes after proof, not before. Walk the ladder to the end — go paper-first rather than straight to live, and let the Risk Gate make the final deterministic admission decision.