Manual backtesting on TradingView

Your strategy is written down. Now the question that matters most: would it have made money? A manual backtest answers that by applying your rules to historical charts, one candle at a time, exactly as if you were trading live. It's slow, and it's honest work — and it's the fastest way to learn more about your strategy in a week than you would in months of live trading.

What you'll learn

• The testing pipeline: in-sample, out-of-sample, and forward testing
• How to prepare a backtest: rules, data range, and recording sheet
• How to run a bar-by-bar backtest without hindsight bias
• What to record for every trade
• How to read the results — and how many trades you need

1. Where backtesting fits

1. In-sample backtest — test (and cautiously refine) your rules on one block of history.
2. Out-of-sample test — freeze the rules and test them on a different block of history you haven't looked at.
3. Forward test — trade the rules on a demo account in real time (the last lesson of this module).

Each stage asks a different question. Skipping stages is how traders go live with strategies that only worked in hindsight.

2. Preparation

Rules. Use your written strategy, with a version number (see Entry, stop, and target rules). If you catch yourself making a decision the rules don't cover, stop and write a rule for it.

Data range. Aim for enough history to include different market conditions — trending and ranging periods, quiet and volatile ones. As a guide:

• Enough to produce at least 100 trades for the in-sample test
• Keep a separate, later block of history untouched for the out-of-sample test

Recording sheet. A spreadsheet with one row per trade:

Recording in R (see Stop-loss / take-profit logic) keeps results independent of account size.

3. Running the test without hindsight

The biggest threat to a manual backtest is hindsight bias — seeing what happened next and letting it influence your decisions.

TradingView's Bar Replay tool helps: you choose a starting point, the chart hides everything after it, and you step forward one candle at a time. (Replay features and available timeframes depend on your TradingView plan; other platforms offer similar tools.)

The process

1. Pick a random starting date in your data range.
2. Step forward candle by candle.
3. At each close, ask only: "Do my rules say to act now?"
4. When a trade triggers, record entry, stop, and target before stepping forward.
5. Step forward until the trade exits according to your rules. Record the result.
6. Repeat until you reach the end of the range.

Include costs. Subtract the typical spread (and commission) from each trade, or record entries at the ask for longs. A strategy that's barely profitable before costs is usually unprofitable after them.

4. Reading the results

Once you have 100+ trades, calculate:

Worked example

(Illustrative backtest of 120 trades.)

• 46 winners averaging +2.1R; 74 losers at −1R
• Win rate: 46 ÷ 120 = 38.3%
• Expectancy: (0.383 × 2.1) − (0.617 × 1) ≈ +0.19R per trade
• Profit factor: (46 × 2.1) ÷ (74 × 1) = 96.6 ÷ 74 ≈ 1.31
• Maximum drawdown: −11R · Longest losing streak: 9 trades

Interpretation: a modest positive edge. At 1% risk per trade, the worst drawdown in the test was about 11% — and live drawdowns are often worse than backtested ones. The trader must be comfortable with 9 losses in a row before going further.

5. Sample size

Small samples lie. With 20 trades, a few lucky winners can make a poor strategy look brilliant — and vice versa.

• Under 30 trades: anecdotes, not evidence
• 100+ trades: a reasonable starting point for a decision
• More, across different conditions: better

If your strategy only produces a handful of trades a year on one market, test it on more markets or a longer history before trusting it.

6. Refining without fooling yourself

It's fine to learn from the in-sample test and improve the rules. But every change you make to fit past data increases the risk of overfitting (next lesson).

• Make few, logical changes with a clear reason ("most losses occurred in the last hour of New York — liquidity is poor then").
• Re-run the in-sample test after each change.
• Then run the out-of-sample test with rules frozen. If results collapse, the changes were probably fitting noise.

Common beginner mistakes

• Hindsight bias — peeking ahead or skipping trades that failed.
• Ignoring costs.
• Drawing conclusions from 20–30 trades.
• Testing only one market condition (for example, a single strong trend).
• Changing rules during the test without restarting it.

Key terms

Practice

1. Set up a recording sheet with the columns above.
2. Choose your data range: an in-sample block for 100+ trades and a later out-of-sample block you won't look at yet.
3. Run the in-sample backtest using Bar Replay (or your platform's equivalent), recording every trade, including costs.
4. Calculate win rate, average win and loss, expectancy, profit factor, maximum drawdown, and the longest losing streak.

Quick recap

• Test in stages: in-sample → out-of-sample → forward test.
• Prepare written rules, enough data for 100+ trades, and a recording sheet in R.
• Use bar replay and record every trade — hindsight bias is the enemy.
• Include costs; judge the strategy on expectancy, profit factor, and drawdown.
• Refine sparingly, then validate on unseen data with rules frozen.