A backtest and a forward test answer different questions — that distinction is covered in more depth in how to backtest a forex robot, which also covers why the gap between the two so often exposes an overfit strategy.
This article picks up exactly where that leaves off: once a strategy’s rules are finalised and the backtest looks solid, how do you actually run a forward test properly, for how long, and how do you know when it’s cleared you to go live?
A forward test is often treated as a formality — a quick demo run before the “real” trading starts. Run carelessly, that’s exactly what it becomes: a box-ticking exercise that catches nothing.
Run properly, it’s one of the few checks that can catch problems a backtest structurally cannot, before any real money is on the line.
What a Forward Test Can Catch That a Backtest Can’t
A backtest, however carefully built, is still a simulation running against historical data with an assumed fill model.
A forward test runs the strategy in real time, against live prices, through the actual execution path it will use going forward. That difference catches several things a backtest simply can’t:
- Execution and platform bugs. Coding errors, broken order logic, or an EA that behaves differently on a live feed than in the strategy tester — these show up in real time, not in historical simulation.
- Real spread and latency behaviour. How your specific broker’s feed actually behaves, not an assumed or historical average.
- Whether the backtest was overfit. A strategy fitted to historical noise has never had to prove itself on data it couldn’t have influenced. A forward test is, by definition, data the strategy’s rules were never shaped around.
- Your own ability to follow the system. Watching a strategy trade in real time, even on demo, surfaces whether the rules are actually followable — or whether you’ll be tempted to intervene once real money is involved.
Setting Up a Forward Test Correctly
- Use a demo account on the same broker, server and account type you intend to trade live. Execution quality, spread behaviour and swap rates all vary between brokers and even between a single broker’s different account types. Testing on a generic or unrelated demo tells you less than you’d think.
- Match the leverage and starting balance to your real plan. A demo account funded with an unrealistic balance, or running different leverage than you’ll actually use, changes position sizing in ways that can mask or exaggerate risk.
- Freeze the strategy’s rules before you start. Whatever version of the strategy finished backtesting is the version that gets forward tested — unchanged. Adjusting rules mid-forward-test because of how a few trades went reintroduces the exact curve-fitting risk a forward test exists to catch.
- Keep a trade log from the first trade. Record entry and exit prices, timestamps, and the reason the strategy took each trade. This is what lets you compare forward-test performance against backtest performance afterwards, rather than relying on impressions.
- Let it run uninterrupted. Resist checking in daily and reacting to short-term swings. The point is to observe the system operating as designed, not to manage it in real time.
How Long Is Long Enough?
There’s no single fixed number of weeks that applies to every strategy, because the right length depends on how often the strategy trades, not how much calendar time has passed.
Two different measures matter, and both need to be satisfied.
- Enough trades. A small number of trades proves very little in either direction. As a rough floor, aim for at least 30-50 trades before drawing any conclusion, and treat 100 or more as a genuinely meaningful sample. A strategy that trades a few times a week may need several months to reach that; one that trades several times a day might reach it within a few weeks.
- Enough market variety. A forward test that happens to run entirely through a single calm, trending month tells you less than one that spans at least one period of higher volatility or choppier conditions. If the forward-test window has been unusually quiet, that’s a reason to extend it rather than conclude early.
A low-frequency swing strategy and a high-frequency scalping system will reach a meaningful sample on very different timelines. Judge by trade count and market variety, not by a fixed number of weeks that happens to feel long enough.
What a Demo Account Still Won’t Tell You
A forward test on demo is a major step up from a backtest, but it isn’t a perfect substitute for live trading, and it’s worth being honest about where it still falls short.
- Demo fills are often more generous than live fills. Some brokers route demo orders differently to live orders, or simply don’t reproduce the same slippage and requotes that occur on a live server during fast moves. A strategy that depends on precise fills can look better on demo than it will perform live.
- There’s no real money, and therefore no real psychological pressure. Following a system’s rules on a demo account, where nothing is actually at stake, is a different experience from following them when a losing streak is costing real money.
- Liquidity and execution during major news events can still differ between a broker’s demo and live servers, even when everything else is matched.
None of this makes forward testing pointless — it still catches genuine bugs, gives a first real read on expectancy, and is a meaningfully closer approximation to live conditions than a backtest. It just isn’t the final word, which is why a small, carefully sized live test is still worth doing before committing full size, even after a clean forward test.
Checklist: Is It Time to Go Live?
| Check | What you’re looking for |
|---|---|
| Sample size | At least 30-50 trades, ideally 100+, not just a few weeks of calendar time |
| Market variety | The test window included more than one type of market condition |
| Consistency with backtest | Win rate, profit factor and drawdown are broadly in line with backtested figures, not wildly different |
| Rule changes during the test | None — the strategy that finished the forward test is the same one that started it |
| Execution issues | No unresolved bugs, missed trades, or unexplained behaviour in the trade log |
| Drawdown behaviour | Any drawdown experienced stayed within, or close to, what the backtest already showed |
A strategy that clears all of these still deserves to start live trading small, with a size you can afford to be wrong about, rather than jumping straight to full position sizing. A forward test reduces uncertainty; it doesn’t eliminate it.
Key Takeaways
A forward test is only useful if it’s run on a matched demo setup, with the strategy’s rules frozen, for long enough to gather a real sample across more than one type of market condition.
Judge it by trade count and consistency with the backtest, not by how many weeks have passed or how good the recent run felt. And remember that even a clean forward test is a strong signal, not a guarantee — moving to live with reduced size is still the safer way to find out what a demo account couldn’t show you.
This article is for educational purposes and does not constitute financial advice. Trading forex carries a high level of risk and may not be suitable for all investors.
Leave a Reply