Forward test

Verification on new data, after backtest, to measure whether edge holds outside the development period.

On this page

Who this is for — Anyone with a method already tested historically who wants to know whether it works on data not "seen" during construction and optimisation.

A forward test evaluates a strategy in a window after the one used in backtest. In practice, you freeze rules and parameters, then observe whether results and behaviour stay coherent.

In plain terms — After the homework exam, you take the in-class test: if the method is real, it does not collapse as soon as the period changes.

Bronze prerequisite — Before this lesson: trading-journal, trade-result, trading-mistake. See bronze-path.


How to read a forward test

A good forward does not perfectly replicate the past, but keeps structure:

  • win rate and payoff in the same "zone" as the backtest;
  • drawdown compatible with planned risk;
  • no dependence on a single lucky period;
  • execution consistent with rules and filters.

If results diverge too much, revisit logic, sample, and operational assumptions.

Example — Backtest over 4 years: expectancy +0.28R. Forward over the next 6 months: +0.21R with similar drawdown. Not identical, but coherent — a plausible edge signal.


Bridge to live trading

Forward test is more credible when combined with:

  1. a window truly separate from the development period;
  2. cost control with slippage-and-fees;
  3. a paper-trading phase to verify routine and discipline.

Card

  • What it is: method test on a subsequent period not used to build it.
  • When to use it: after backtest and before increasing live size.
  • Typical mistake: changing rules during the test and still calling it forward.

Silver path — Module: Validation. Part of silver-path.