Forward testing and going live
A good backtest proves that a strategy would have worked. Forward testing checks whether it does work in real time — with live data, real spreads, real delays and your real behaviour. It is the bridge between research and real money, and it catches problems that no backtest can reveal.
What forward testing reveals
| Issue | Why a backtest misses it |
|---|---|
| Execution delays and slippage | Backtests often assume perfect fills |
| Spread changes | Historical data may use fixed or average spreads |
| Missed signals | You may be asleep, at work or distracted |
| Emotional hesitation | No emotions in a backtest |
| Coding errors | Logic may behave differently in live data |
| Changed market conditions | The future may differ from the test period |
Types of forward testing
- Paper trading: record trades manually in real time without placing orders. Cheapest, but ignores execution.
- Demo account: place real orders on a broker's demo server. Tests execution and platform, but demo fills may be better than live.
- Micro live: trade the minimum size (0.01 lot) with real money. The most realistic test, with very small risk.
Many traders use all three in sequence. See from demo to live.
How long to forward test
Long enough to collect a meaningful sample. A strategy taking 10 trades a month needs several months for 30–50 trades. A rough guide:
- Minimum 30 trades before drawing conclusions.
- Ideally 50–100 trades, covering different market conditions.
- For low-frequency strategies (a few trades a month), consider running several instruments to gather data faster.
Comparing forward and backtest results
| Metric | Backtest | Forward test | Acceptable? |
|---|---|---|---|
| Win rate | 46% | 43% | Yes — within normal variation |
| Average win | 2.0R | 1.8R | Yes — slight slippage |
| Expectancy | +0.38R | +0.20R | Yes, but monitor |
| Rule breaks | 0 | 12% | Needs work |
Expect forward results to be somewhat weaker than the backtest. If expectancy drops close to zero or below after a decent sample, investigate before going further. Calculate with the expectancy calculator.
Is the difference just luck?
Small samples vary a lot. With a 45% win rate, 30 trades could easily show anywhere from about 30% to 60% wins purely by chance. Before abandoning a strategy, check:
- Did I follow the rules exactly?
- Is the difference larger than normal random variation for this sample size?
- Have market conditions changed (for example, volatility much lower)?
Keep a forward-test log
For each signal, record in your journal:
- Signal time and planned entry.
- Actual entry and slippage.
- Whether you took it (and why not, if skipped).
- Result in R, and MAE/MFE.
- Notes on news, session and emotion.
Skipped signals are important data: if the skipped trades were mostly winners, your hesitation is costing you.
Automated strategies
For expert advisors or bots, run them on a demo account or a virtual private server (VPS) for at least one to three months. Check that orders, stops and targets are placed as intended, especially around rollover and weekends. Compare each trade with what the backtest would have done on the same data.
From forward test to live
- Start at micro size or 0.25–0.5% risk.
- Increase gradually after 30–50 live trades that match expectations — see scaling up.
- Keep a stop rule: if live drawdown exceeds the backtest's worst drawdown by a set margin (for example 1.5×), pause and review.
Frequently asked questions
Can I skip forward testing if my backtest is strong? It's risky. Many strategies that look great in backtests fail on execution details, which forward testing reveals cheaply.
Should I change rules during the forward test? No. Changing rules mid-test makes the results meaningless. Note ideas and test them separately later.
Is a demo account enough for forward testing? It is a good start, but demo fills can be better than live ones. A short period of micro-size live trading shows the real execution quality.
Key takeaways
- Forward testing checks execution, slippage, missed signals and behaviour.
- Use paper, demo and micro-live stages; collect at least 30–50 trades.
- Expect slightly weaker results than the backtest; investigate large gaps.
- Log skipped signals and don't change rules mid-test.
Figures are illustrative. Education only, not investment advice.
Educational content by CROZAX. Not financial advice. Trading involves risk.