Forward test (incubation)
A forward test runs a frozen trading rule on market data that arrives after the rule was designed, usually on a simulated or paper account, before any capital is committed. It is the only out-of-sample evidence that no design decision could have seen, because the data did not exist when the rule was written.
Senzoukria · Glossary · Updated September 2026
Forward test versus historical out-of-sample
A historical out-of-sample period was, in principle, set aside. In practice it has often been looked at, charted or discussed, and decisions leak into it. A forward test removes that possibility: the rule, its parameters, the instrument and the costs are written down on a date, and only data after that date counts. The period is often called incubation.
It also tests what a backtest cannot: the live data path, the latency of decisions, the discipline of executing every signal and the realized slippage.
What to compare, and after how long
Compare the distribution of forward results with the backtest, not a single number. Suppose the backtest shows an average of +25 dollars per trade with a standard deviation of 180. After 40 forward trades, the standard error of the average is 180 ÷ √40 ≈ 28.5. A forward average of −5 is then about one standard error below the backtest, which is not yet evidence of failure. The same shortfall after 400 trades, with a standard error near 9, would be.
Set the length of the incubation and the rejection criterion in advance, for example a number of trades and an average below a stated threshold. Otherwise the decision to stop is taken on the most recent streak.
- Record every signal, including those not taken, with the reason.
- Compare realized slippage and fees with the backtest assumptions.
- Check that the forward trade frequency matches the backtest; a large difference points to a data or rule mismatch.
In Senzoukria
Three environments can host a forward test without a funded account. The Auto strategy panel runs a saved strategy script on the local simulated account, deciding once per closed bar, and its source excludes any route to a broker; the demo account's own limits, such as a maximum daily loss and a contract cap, apply before each entry. A broker paper account on a paper plant accepts real orders from the order ticket with simulated fills. And the autopilot, which can send orders to a broker account, does nothing until a person arms it explicitly and runs under its own maximum contracts and maximum daily loss. The journal then holds the trades for comparison with the backtest statistics.
Related
- In-sample vs out-of-sample
- Paper trading account
- Simulated account
- Running a strategy on the simulated account
- Autopilot
In the same section
- Forward volatility
- Footprint vs candlestick
- Front month
- Footprint software
- Funded account
- Footprint chart
- Funding rate
- Cell types
This page in other languages
Frequently asked questions
- How long should a forward test last?
- Long enough to collect a number of trades that can distinguish the backtest average from zero or from a clearly worse value. For low-frequency strategies that can mean months. Fix the count or the duration before starting.
- Is paper trading a valid forward test?
- It is valid for the rule's signals and for the trader's execution discipline. Its fills are simulated, so realized slippage on a funded account may differ, especially for limit orders at turning points and for stops in fast markets.