How to Backtest a Trading Strategy Manually: A Step by Step Guide

Write your rules down, then replay the chart one candle at a time from a start date you didn’t pick by eye. Take every trade the rules give you and record each one. Stop only when you have enough trades to separate an edge from luck: usually several hundred, not 100. Then compute expectancy, profit factor and drawdown from the log.

The steps are simple. The discipline around them is what makes the result trustworthy. That means hiding the future, recording every trade with the ugly ones included, and reading the result with the right sample size. This guide covers each step and the mistakes that quietly ruin a backtest.

What manual backtesting is, and when you need it

Manual backtesting means replaying historical price and trading it by hand, as if it were live. Automated backtesting means writing the strategy as code and letting a computer run it across the history.

If your strategy can be written as exact rules, automate it: a computer will test thousands of trades without fatigue or bias. If it depends on judgment, such as reading structure or whether a setup “looks right”, manual is the only way to test it. The judgment is part of the strategy. Manual testing also trains you on hundreds of instances of your own setup. Demo trading delivers them only as fast as the market forms them.

Step 1: Write the rules down

If you can’t write it, you can’t test it. Before opening a chart, write:

• Market and timeframe, including any higher timeframe you use for context
• Entry conditions, each one specific enough that someone else could apply it
• Stop loss: where, and why there
• Take profit or exit rule: fixed target, trailing, or a condition
• Filters: sessions you trade, news you avoid, days you skip
• Risk per trade, as a percentage of the account or in R

Turn the entry conditions into a checklist you can tick for each trade. Later, the checklist answers a question most backtests can’t. Did the trades that met every condition do better than the ones where you bent a rule?

Step 2: Choose the period

One year of one quiet market proves very little. Pick a period that includes different conditions: trending and ranging, high and low volatility, and at least one period of major news.

Be realistic about scale. A single year of forex contains roughly:

Timeframe	Candles in one year
Daily	260
4 hour	1,560
1 hour	6,240
15 minute	24,960
5 minute	74,880

Counted as 52 weeks of 5 trading days, 24 hours a day, on the 17:00 New York grid. Holidays remove a few.

Lower timeframes produce more trades per year, which helps with sample size, but every candle has to be watched. Plan the work accordingly: a 15 minute strategy tested over two years means replaying about 50,000 candles.

In BacktestFlow, forex history starts between 2010 and 2015 depending on pair, which gives up to 16 years on the majors.

Step 3: Choose the tool

We make BacktestFlow, one of the tools below.

You need three things: a replay that hides the future, fills at the prices the chart showed, and a log you can compute from.

• A charting tool plus a spreadsheet. TradingView Bar Replay replays the chart well, but it records nothing. Per its documentation, orders during replay execute on real time data. So every entry, exit and result goes into the spreadsheet by hand.
• A dedicated replay backtester. These fill orders on the replayed candles and keep the log for you. FX Replay goes wide across many markets, with seconds data. Forex Tester Online leads on tick data. BacktestFlow is manual backtesting software built for forex. Its replay hides everything after the current candle and waits until each close is saved. We compare ten tools in the best forex backtesting software.

The spreadsheet route works. Its hidden cost is the trades that never make it into the sheet. Leave out the losers you’d rather forget, and the backtest flatters the strategy.

Step 4: Hide the future

A manual backtest is only as good as your ignorance of what happens next. The rules:

• Pick start dates you don’t choose by looking at the chart. Random dates, or the first trading day of each month. A start date picked because the chart “looks interesting” is already biased.
• Never scroll right. Only the candle the replay has reached exists.
• Be careful switching timeframes. A higher timeframe candle that hasn’t closed can show you where price goes next. Check how your tool builds it. In BacktestFlow, switching timeframe rebuilds the current candle only from bars already replayed.
• Don’t test on a period you remember trading. If you remember what EURUSD did in a given week, test a different week.

Step 5: Trade it, and record every trade

Step the chart forward. When the rules give you a setup, take it, even if you don’t like how it looks. Skipping setups you dislike is the fastest way to turn a backtest into a highlight reel.

Record every trade with at least these ten columns:

Column	Example
Date and time	2024-03-12 09:15
Pair and direction	EURUSD, long
Entry, stop, target	1.0912, 1.0897, 1.0942
Risk in R	1R (15 pips)
Result in R	+2R, or −1R
Session	London
Checklist items met	5 of 5
Rule broken, if any	Entered before candle close
Screenshot	link
Notes	Clean retest after news

Record the result in R (multiples of the amount risked) rather than money or pips. R makes trades comparable across different stop sizes. It also lets you change position sizing later without redoing the test.

Also record setups you saw but didn’t take, with the reason. If the skipped ones would have won more often than the taken ones, that is worth knowing. In BacktestFlow, a setup you didn’t take can be saved as a missed trade. It stays in the record as a case study, reported as the cost of missed opportunities.

Step 6: Get to a sample that means something

This is where many manual backtests stop too early. The common “test 100 trades” rule has no statistical basis. To tell a genuine 55% win rate from a 50% coin flip at 95% confidence takes about 385 trades. A stronger edge needs fewer; a weaker one needs more.

The practical rule: keep adding trades until the confidence interval around your result no longer includes break even. Break even means zero expectancy, or a 50% win rate at 1:1 reward to risk. We worked through the math, with a table for different edges, in how many trades you need to backtest a strategy. You can check any figure with the sample size calculator.

Step 7: Compute the statistics

From the log, compute at least these four:

• Expectancy: the average result per trade, in R. Win rate × average win − loss rate × average loss.
• Profit factor: gross profit divided by gross loss. Above 1 means the strategy made money.
• Maximum drawdown: the largest fall from a peak in the running total, and how many trades it took to recover.
• Longest losing streak: the run you have to be able to sit through live.

Expectancy matters more than win rate. Compare two strategies (or put in your own numbers with the expectancy calculator):

	Strategy A	Strategy B
Win rate	40%	70%
Average win	2.5R	0.5R
Average loss	1R	1R
Expectancy per trade	+0.40R	+0.05R

Strategy B wins far more often and makes an eighth as much per trade. Its edge is thin enough that spreads and a few execution mistakes would erase it.

Step 8: Break it down, carefully

Once the overall result holds up, split it: by session, by pair, by day of week, by whether every checklist item was met. The splits often show where the edge actually lives. A strategy might work in the London session and lose in Asia. Your best time to trade forex is something you measure, not a universal rule.

Be careful, though. Every split divides the sample. Sixty trades across five sessions leaves twelve per session. Twelve trades will show convincing differences purely by chance. Only act on a split with enough trades in each part to stand on its own.

BacktestFlow’s Flow Coach applies this rule for you. A split counts as a finding only if it passes a significance check: a 90% Wilson interval and at least 8 trades per group.

Step 9: Forward test before trading it live

A backtest shows what the rules would have done. It can’t show how you will execute them in real time, with real spreads and your own hesitation. Run the strategy on a demo account or at very small size for a few weeks. Then compare the results with the backtest.

If live results fall well short of the backtest, the cause is often execution rather than the strategy: late entries, moved stops, skipped setups. We cover how to tell the two apart in why backtest results don’t match live trading. BacktestFlow keeps a live journal separate from your backtests, with its own statistics, so you can compare the two.

If the goal is a prop firm challenge, test the strategy against the firm’s drawdown rules before you pay. We show how in how to backtest a prop firm challenge. On the Pro and Max plans, BacktestFlow runs a challenge’s profit target and drawdown limits as a simulation on historical candles. BacktestFlow is not affiliated with any prop firm.

Mistakes that quietly ruin a manual backtest

• Skipping setups that look bad. Take every trade the rules give you.
• Picking start dates by eye, which selects the periods where the strategy looks good.
• Changing the rules mid test. If you change them, start the count again.
• Stopping at 30 or 100 trades and treating the win rate as proven.
• Recording in money instead of R, which makes trades with different stops impossible to compare.
• Testing only one market regime. A trend strategy tested only on trending months will look excellent.
• Not recording broken rules. The trades where you broke a rule are the ones that explain the gap between backtest and live results.

A short checklist

1. Rules written down, entry conditions as a checklist
2. A period covering different conditions
3. A tool that hides the future and fills at historical prices
4. Start dates not chosen by eye
5. Every trade recorded in R, including skipped setups and broken rules
6. Enough trades for the confidence interval to clear break even
7. Expectancy, profit factor, drawdown and longest losing streak computed
8. Splits only where each part has enough trades
9. A forward test before real money

Nothing here is financial advice. Backtested performance does not guarantee future results.