Knowledge baseFeatures

Test Lab & Proof: validation is research, not gambling

Backtest a system, stress it with walk-forward, compare it against buy-and-hold, and read the proof ladder that says which rungs still stand between it and capital. Weaknesses surface before capital does.

5 min read

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.

Reading is free. So is testing.

Paper trading costs nothing and needs no card. Build a system, backtest it on real candles, and decide from the result.