Why Most Custom Indicators Look Good but Trade Poorly
Custom indicators often look amazing on a chart but trade poorly because you're seeing them with the benefit of hindsight. When you scroll back, every signal lines up with a move that already happened, which feels like proof. The problem is usually repainting and lookahead: some indicators quietly redraw their past signals or use information that wasn't available yet, so live they behave very differently. On top of that, our eyes are drawn to the times the indicator was right and skip the times it wasn't. A tool tuned to fit past price will hug it beautifully and then miss on new data, which is called overfitting. Add real costs and the gap between a clean chart and messy live execution, and the pretty picture fades. Indicators give context, they don't guarantee the trades will follow through.
Why an indicator that appears to identify setups visually often fails to produce results in a structured backtest or live trading context.
Key points
- On a chart you see signals with hindsight, so every past marker seems to sit right before a move that already happened.
- Repainting is a big culprit: some indicators redraw their past signals so history looks cleaner than it was in real time.
- Lookahead bias sneaks in when a script uses data that wasn't actually available at that moment on the chart.
- Your eyes naturally focus on the times the indicator was right and skip the times it failed, which flatters it.
- An indicator tuned to fit past price hugs it perfectly and then stumbles on new, unseen data, which is overfitting.
- Real costs and messy live execution widen the gap between a pretty backtest and actual results.
Frequently asked questions
What is repainting and how do I spot it?
Repainting is when past signals change after they first appear, so the history looks better than it was live. You can spot it by watching how the indicator behaves as new bars form, or by using bar replay to see whether old markers move.
What is lookahead bias in an indicator?
Lookahead bias is when a script uses information that wasn't available yet at that point in time, like a value from a higher timeframe that hadn't closed. It makes the past look predictable when it wasn't, and it fades the moment you trade live.
How can I tell if an indicator is overfit?
If it fits past price almost perfectly but falls apart on new data, a different symbol, or a different period, it's likely overfit. A tool built to hug one stretch of history rarely holds up once the market changes.
Why do indicator backtests overstate performance?
They often ignore trading costs, benefit from hindsight, and can include repaint or lookahead effects that wouldn't exist live. Together these make the reported result look far better than what you'd actually experience.
How can I test an indicator's idea honestly?
In Agenticks, AlgoAgent can turn the idea into a testable strategy and backtest it on past data, then show you the results including drawdown, so you see how it really behaves rather than how it looks. It's research to inform you, not a guarantee.
Related on Agenticks
This content is for educational purposes only and does not constitute financial advice. Trading involves risk, including possible loss of capital.