Forward testing means trading your strategy, with its rules fixed, on live market data that did not exist when you built it. You do it on a demo account first, then on a very small live account. It tells you what a backtest cannot: whether your rules still work on prices you have never seen, and whether you can follow them when real candles are moving.
Below is the plan we use with students: what to set up, what to record, how long to run it, and how to read the result without fooling yourself.
TL;DR
- Backtesting uses old data. Forward testing uses new data as it arrives, so it is the check against curve-fitting.
- Freeze your rules in writing before the first trade. Changing them mid-test resets the count.
- Log every trade with a screenshot, the R result and a note on whether you followed the plan.
- Aim for at least 50 trades across different market conditions, then move to a small live account only if the numbers and your behaviour both hold up.
- A failed forward test is cheap information. A failed live account is not.
Table of Contents
- What forward testing is, and how it differs from backtesting
- Before you start: freeze the plan
- Stage one: forward test on a demo account
- What to measure
- How long to run it and how many trades you need
- Stage two: a small live account
- Reading the results: continue, fix or stop
- Common mistakes that spoil a forward test
- Perspective: why we put forward testing before everything else
- Want help running your test?
- FAQ
- Sources
What forward testing is, and how it differs from backtesting
A backtest replays your rules on historical charts. A forward test applies the same rules going forward, one trade at a time, with no knowledge of what comes next. CMC Markets’ UK guide describes forward testing as using new data the strategy has not seen, while backtesting uses historical data. It also notes the trade-off: backtesting is fast but can overfit the past, while forward testing validates a strategy on fresh conditions but means waiting for new data.
That waiting is the point. When you backtest, you already know how the story ends, and it is very easy to nudge a rule here and a filter there until the equity curve looks great. That is curve-fitting. A forward test removes the option. The market gives you the next candle and you either follow your rules or you don’t.
One quick clarification, because the term trips people up. In currency markets, a “forward” is also a contract that fixes an exchange rate for a future date. That is a different subject from testing a strategy, and spot versus forward rates explains the pricing side if you are curious. Everything below is about testing your trading rules.
| Backtesting | Forward testing | |
|---|---|---|
| Data | Historical, already known | New, arriving live |
| Speed | Hours or days | Weeks or months |
| Main risk | Overfitting the past | Too few trades, or bending the rules |
| What it proves | The idea had an edge once | The edge survives, and you can execute it |
| Cost | Free | Free on demo, small on live |
They are two stages of one process, not rivals. If you have not done the first stage properly yet, start with our guide to backtesting a forex strategy and come back.
Before you start: freeze the plan
A forward test only counts if the rules cannot move. Write a one-page document and do not edit it until the test ends. It should cover:
- Market and timeframe. One pair, one execution timeframe. EUR/USD on the 1-hour chart is a sensible start because spreads are tight.
- Session. When you will trade. For UK traders, the London session and the London and New York overlap are the usual windows. Pick yours and stick to it.
- Entry rules. Exactly what must be on the chart before you click. If two people could read your rule differently, it is not a rule yet.
- Stop-loss and target rules. Where the stop goes and how the target or exit is decided.
- Risk per trade. A fixed percentage of the account. Most of our students use 1%, and our piece on the 2% risk rule shows how to turn that into a lot size in cash terms.
- Stop conditions. A daily loss limit, and the number of trades after which you review.
- What counts as a pass. Decide now: minimum expectancy, maximum drawdown, minimum plan-following rate.
That last item matters most. If you set the pass mark after seeing the results, you will set it just below whatever you got.
If you are still choosing a strategy to test, our rule-based breakout retest guide is a good example of rules written tightly enough to forward test.
Stage one: forward test on a demo account
Demo is where you find out whether the rules run in real time. It costs nothing, so use it to find the obvious problems.
Set the demo up like the real thing
A demo that is easier than live teaches you the wrong lessons. Set the balance to the amount you would actually fund, not the default six-figure sum. Use the same broker platform you intend to trade on, so you see real spreads and the same order types. Risk the same percentage you plan to risk live. Our demo account beginner’s guide walks through the setup.
Be honest about what demo cannot show. Fills are cleaner, there is no real slippage in fast news, and nothing is at stake, so fear and greed are muted. That is why demo is stage one and not the whole test.
Record every trade
Open a spreadsheet or journal before the first trade. For each one, log:
- Date, time and pair
- Setup name and the rule that triggered it
- Entry, stop, target and position size
- Screenshot at entry and at exit
- Result in R (your risk unit), not just pips or pounds
- Followed the plan? Yes or no, and if no, what you did instead
- One line on how you felt
The “followed the plan” column is the one most people skip, and it is the most revealing. A strategy that wins 55% of the time when you follow it and loses when you improvise does not have a strategy problem. It has an execution problem, and you need to know which one you are dealing with.
If you already keep a journal, use it. If not, our guide to trading performance review covers what to track and how often to review.
Do not trade what you did not plan
If a setup appears that is not in your document, you can note it as an observation, but it does not go into the test results. Mixing improvised trades into the sample makes the numbers meaningless.
What to measure
A total profit figure on its own tells you very little. Look at these:
| Metric | What it tells you | How to work it out |
|---|---|---|
| Win rate | How often you are right | Winners divided by total trades |
| Average win and average loss in R | Whether winners pay for losers | Mean R of each group |
| Expectancy | Average R per trade | (Win rate × average win) minus (loss rate × average loss) |
| Maximum drawdown | The worst losing stretch | Peak to trough in R or % |
| Longest losing streak | What you must be able to live with | Count in the log |
| Plan-following rate | Execution discipline | Trades that followed the plan divided by all trades |
| Costs | What spread and commission take | Compare R before and after costs |
A worked example. Over 50 trades you win 22 and lose 28. Your average win is 2R and your average loss is 1R. Win rate is 44%. Expectancy is (0.44 × 2) minus (0.56 × 1), which is 0.88 minus 0.56, so 0.32R per trade. Risking 1% per trade, that is roughly 0.32% of the account per trade on average, before you have allowed for the noise in a sample this small. A 44% win rate looks poor, yet the system is positive because the winners are twice the size of the losers.
Now check behaviour. Did the longest losing streak make you skip the next signal? Did you move a stop? Those are the results that decide whether you are ready for live money.
How long to run it and how many trades you need
There is no official number, and anyone who gives you one as a law is guessing. What we can say from practice is this: the sample has to be big enough to survive a bad run and varied enough to include different conditions.
- Under 30 trades is a smoke test. It can reveal that a rule is unworkable. It cannot prove an edge.
- 30 trades is a reasonable first checkpoint. Our 30 demo trades on EUR/USD walkthrough shows what that looks like at 1% risk.
- 50 trades or more is where we would start to trust the averages, and it is the number in our forex learning path.
- Time matters as well as count. Fifty trades in one strong trending fortnight tell you about that fortnight. Try to cover trending weeks, ranging weeks and at least one high-impact news period.
For a swing or daily-chart strategy that trades a few times a month, this can take several months. That is normal. Do not shorten it by loosening your entry rules to get more trades.
Remember that even 50 trades is a small sample statistically. Treat the result as evidence that raises or lowers your confidence, not as a guarantee. A good 50 trades means “keep going with more care”, not “this will keep winning”.
Stage two: a small live account
Demo cannot test your nerves. A small live account can, if the amount is small enough that a loss does not hurt your life but large enough that you feel it.
Before you fund it, understand the odds you are up against. The FCA requires firms offering CFDs to retail clients to show a standard warning, including the share of retail accounts that lose money with that provider, in its rules on CFD marketing and sale. Read that percentage on your broker’s site. It is a good reminder of why the live stage should start small.
How to run it:
- Use the same rules, the same risk percentage and the same log. Nothing changes except that the money is real.
- Keep position sizes tiny at first. If your demo risk was 1%, you can still use 1% of a small balance.
- Run another 30 to 50 trades.
- Compare the live log to the demo log, line by line: expectancy, drawdown, plan-following rate, costs.
Expect live to be slightly worse. Spreads, slippage and emotion all take a bite. The question is whether it is slightly worse or completely different. If the plan-following rate drops sharply once money is on the line, you have just learned the most useful thing the test could teach. Our guide to validating a forex strategy in live markets goes deeper on this stage.
Reading the results: continue, fix or stop
Use the pass mark you wrote down before starting. Then take one of three paths.
Continue to scale up. Expectancy is positive after costs, drawdown stayed inside your limit, and you followed the plan on at least 90% of trades. Increase size slowly, in small steps, and keep logging.
Keep testing. Results are positive but the sample is small, covers one market condition, or the plan-following rate is shaky. Extend the test. Do not go live yet.
Stop and rebuild. Expectancy is flat or negative, or the live results diverge widely from the backtest. Go back to the rules, change one thing, and start a new forward test from zero. Never patch a rule mid-test and keep the old trade count.
The temptation after a losing streak is to edit the rules to avoid the pain. That is a revenge trade in slow motion: the brain stops trading the chart and starts trading the pain. A hard rule helps here: no rule changes until the pre-agreed trade count is reached.
Common mistakes that spoil a forward test
- Cherry-picking. Logging only the trades you liked. Log everything, including the ones you skipped by mistake.
- Oversized demo risk. Risking 10% on demo because “it isn’t real” creates habits and results that mean nothing.
- Changing the rules after a loss. This turns the forward test back into curve-fitting, only slower.
- Ignoring costs. A scalping system that looks fine before spread can be negative after it.
- Testing too many ideas at once. Run one strategy at a time or the data gets muddy.
- Stopping at the first good week. A hot streak is not a result.
Perspective: why we put forward testing before everything else
We have spent more than 18 years trading live markets, and the pattern we see in struggling traders is rarely a lack of strategies. It is a lack of evidence. They jump from one setup to the next after a few losses, never collecting enough trades to know whether anything works.
Forward testing fixes that. It forces you to pick a plan, hold it still and let the market judge it. It also shows you something no backtest does: how you behave when a plan hits a losing run. Sounds strange, but the most valuable number in many students’ logs is not expectancy. It is the plan-following rate.
Our advice is simple. Be slow at this stage and fast at nothing. A few months of careful logging will save you from years of repeated guesses. If you lose during the test, that is the test doing its job, at a price you chose.
— Gabriel
Want help running your test?
If you would rather not do this alone, we teach this process inside our courses and mentorship. The Price Action (Core) plan is £149 per month, and a single mentorship session, where we review your log and rules together, is €149. You can see every plan, including the 3-session and 5-session mentorship packages, on the Trader Gibkey site.
FAQ
Is forward testing the same as paper trading?
Mostly, yes. Paper trading on a demo account is the first stage of forward testing. A full forward test adds fixed written rules, a complete trade log and a small live stage afterwards.
How long should I forward test before going live?
Until you reach your pre-set trade count, ideally 50 or more trades across different market conditions. That can take weeks for an active strategy and months for a slower one. Time alone is not the measure, the sample is.
Can forward testing guarantee I will profit?
No. It reduces the chance of trading a curve-fitted strategy, but markets change and a small sample can mislead. Trade small, keep logging, and only risk money you can afford to lose.
Sources
- CMC Markets: backtesting strategies on TradingView, UK guide
- FCA Handbook, COBS 22.5: restrictions on retail CFD marketing, distribution and sale
- Fiatmap: spot vs forward rate