1. Start with a trade idea, not a prediction
A trade idea is testable. "EUR/USD will go up" is not a strategy. "After a London session liquidity sweep below Asian range low, enter long on a bullish close back inside the range, stop below the sweep, target 2R or the opposite range high" is testable.
The more exact the rule, the easier it is to judge. Precision does not mean the strategy must be mechanical forever. It means the backtest has a fair boundary.
2. Choose the market, timeframe, and session
Forex behavior changes by pair and time of day. GBP/USD during London open is not the same as USD/CAD during late New York. Choose one or two markets first. Keep the timeframe consistent. If your entry is on M15 and your bias is on H1, write that down.
For many traders, session filters are where the truth appears. A setup may work during London and become random during dead hours. A proper backtest separates those conditions instead of mixing them into one blurry average.
3. Define entry, stop loss, and target rules
Your entry rule decides when the trade starts. Your stop loss decides where the idea is wrong. Your target rule decides how profit is taken. If any of these are vague, the results will be vague.
Common target models include fixed 1:2 risk-to-reward, previous high or low, next liquidity pool, session range edge, ATR multiple, or partial closes. There is no universal best choice. The right question is whether the target logic matches the reason for the trade.
4. Decide what counts as no trade
No-trade rules are as important as entry rules. Many strategies are damaged by taking them in the wrong conditions. Examples include high-impact news, very wide spreads, end-of-day illiquidity, overlapping major releases, or market structure that is too compressed to offer clean risk.
Write the filter before the test. If you add it after seeing the loss, you are optimizing the past instead of learning from it.
5. Use consistent risk per trade
Use one risk model during the test. Many traders choose 0.5 percent or 1 percent per trade for analysis. The number is less important than consistency. If trade one risks 0.25 percent and trade two risks 3 percent because it "looked better", your analytics will not explain the strategy. They will explain your confidence swings.
Backtest discipline: do not increase risk after winners or reduce risk after losses inside the same sample. That adds psychology into a strategy test.
6. Replay and log every valid setup
Move candle by candle. When the setup appears, take it or reject it based on the written rules. Record the trade immediately. Do not skip a valid setup because you dislike how it later ended. Those skipped trades are usually where the backtest becomes dishonest.
Your log should include result in R, pips, money, session, direction, setup tag, and short notes. A strategy with positive expectancy but terrible execution notes may still need work before live trading.
7. Review expectancy, not only win rate
Win rate alone is incomplete. A 35 percent win rate can be profitable if winners are large enough. A 70 percent win rate can lose money if losses are much larger than winners. Expectancy combines win rate, average win, and average loss into the average result per trade.
Use this simple logic: if the average trade is positive after costs, the idea may be worth more testing. If the average trade is negative, either the rules need revision or the market condition does not fit the setup. How to calculate expectancy and read the other numbers is covered in the analytics guide.
8. Check drawdown and losing streaks
A strategy can be profitable and still emotionally impossible for you to trade. If the backtest shows a 12-trade losing streak, you need to know that before going live. Drawdown tells you how deep the account dipped from its prior level. Losing streaks tell you what your discipline will have to survive.
Many traders abandon good systems because they never tested the ugly part. A proper forex backtest includes the ugly part on purpose.
9. Avoid curve fitting
Curve fitting happens when you keep adjusting rules until the historical sample looks perfect. You remove one hour, change one target, add one indicator, and suddenly the past looks amazing. The problem is that the new rule may only fit those exact candles.
A healthy workflow is to develop rules on one sample, then test the final version on a fresh out-of-sample period. If performance collapses on fresh data, the rules were probably too tailored to the first sample.
10. Decide the next action
At the end, make one of four decisions: discard the idea, revise one rule and retest, forward test on demo, or trade live with small risk. Do not jump from one good backtest week to full-size live trading. The backtest is evidence, not a guarantee.
Proper backtesting FAQ
Should I backtest one pair or many pairs?
Start with one pair so the rules are clean. After that, test related pairs separately. Do not combine all pairs into one result until you know each market behaves acceptably.
What is out-of-sample testing?
It means testing the finished rules on historical data that was not used while building the strategy. It helps reveal whether the idea is robust or just fitted to the first sample.