Automation
Reading the TradingView Strategy Backtest Report: Spotting Overfitting
Between "I feel this makes money" and "the data says it made money over five years" sits one honest backtest. Swap indicator() for strategy(), add entry/exit orders, and the tester at the bottom auto-generates a report.
Read these four numbers first
- Max drawdown: look at it before net profit — it decides whether you can hold the system long enough to reach profit;
- Payoff × win rate: only the product is the expectancy; either alone will fool you;
- Number of trades: under 100 has essentially no statistical meaning;
- Commission and slippage: fill them honestly; omitting them on an intraday strategy is self-deception.
Three signs of overfitting
- Nudge a parameter and performance changes a lot — you optimized noise;
- It only shines on one instrument / one period — switch both to check;
- More and more rules, each to "fix" a few specific losses — that's memorizing the data.
Out-of-sample: tune on 2020–2024, validate on 2025-to-now; if performance collapses, start over. A backtest that skips this is just a pretty curve. To write a strategy, see Pine Lesson 1.