How to backtest a trading strategy

You cannot backtest a vague idea like "buy when the trend looks strong." A strategy you can test has rules with no room for opinion. Be specific about four things. One, the setup: the exact condition that puts a trade on your radar, for example a pullback on EUR/USD during the London session. Two, the entry: the precise price or signal that gets you in. Three, the stop loss: the price where you admit the idea was wrong and exit, which defines your risk. Four, the take profit or exit rule: where you get out in profit, or the condition that closes the trade. Also fix your timeframe (say the 1-hour chart), the pair, and the session hours in UTC. If two people reading your rules would place different trades, the rules are not finished. The whole point of backtesting is to remove your in-the-moment guessing, and you can only do that if the rules decide for you.

There are two honest ways to do this. Manual backtesting means loading historical charts (most trading platforms let you scroll back through past price), moving forward one candle at a time, and logging every trade your rules would have taken. It is slow, but you learn the most because you see real price behaviour. Automated backtesting means coding your rules into software (platforms like TradingView's Pine Script, MetaTrader's Strategy Tester, or Python libraries) so a computer runs them over years of data in seconds. It is fast and removes some human bias, but it only tests exactly what you coded, and a small bug can make results meaningless. A common honest path is to test manually first to understand the strategy, then automate once the rules are stable. Whichever you pick, you need enough history. A few weeks proves nothing. Aim for a sample that covers different conditions: trending markets, quiet ranges, and high-volatility news periods.

For each trade, write down the date and time in UTC, entry price, stop, exit, and the result in "R," meaning multiples of the amount you risked. If you risked 20 pips and made 40, that is +2R; if you got stopped out, that is -1R. Using R instead of money keeps the test about the strategy, not your account size. After 50 to 100 trades you can look at a few honest numbers. Win rate is the share of trades that won, but it means little on its own. A strategy can win 40 percent of the time and still hold up if winners are bigger than losers, and a 70 percent win rate can still bleed if the losers are huge. Pair win rate with average win versus average loss. Also look at the largest losing streak and the deepest drawdown, the biggest peak-to-valley drop in your running total, because that is the pain you would actually have to sit through. Be brutally honest: count every losing trade, do not quietly skip the ones that hurt.

This is where most beginners fool themselves. The past does not repeat exactly, so good historical results are not a promise of anything. Several traps make backtests look better than reality. Hindsight bias: when you already know how the chart played out, it is easy to "see" the perfect entry you would never have taken live. Curve fitting (also called over-optimisation): tweaking rules until they fit old data perfectly, which usually means they break on new data. Ignoring costs: real trades pay the spread (the gap between the buy and sell price) and sometimes commission and slippage, so a backtest with zero costs is fiction. Survivorship in your memory: remembering the wins and forgetting the chop. And the obvious one: clicking a button in a calm test is not the same as holding a losing trade with real money on the line. Trading is risky and most retail traders lose money, and a clean backtest does not change that. Treat a backtest as a filter that kills bad ideas, not proof that a good idea will work.

A strategy that survived backtesting has earned one thing: a forward test. Forward testing (or paper trading) means running the exact same rules on live, unfolding price using a demo account, where you do not know what happens next. This catches the problems backtesting hides, because you have to make decisions in real time without seeing the future. Run it for a meaningful stretch, weeks at least, and log trades the same way you did in the backtest. Compare the two. If forward results fall apart, the backtest was probably curve fitted, or you were unconsciously cheating on the history. Only after a strategy holds up both backward and forward should real money even be a conversation, and even then only with money you can afford to lose. The discipline you build here (writing rules, recording results, sitting with losing streaks) is the part that transfers to any market, but the testing itself should stay focused on the forex pairs and timeframes you actually plan to trade.