I tried to make my Pine Script indicator beat my discretion. It didn't.

I have a setup I trade by eye: Keltner Zones with a volatility squeeze filter and a price-action confirmation. It works for me. So I did the obvious experiment — I coded the whole thing in Pine Script v5 with clean parameters, ran it on three years of data on EURUSD and US100, and watched the equity curve climb.

I switched it on for live trading. Three months. It made money, but less than my eye, and with a worse drawdown profile.

Why the indicator lost

Markets change regimes. The indicator does not. The squeeze filter that works in a ranging environment hands you whipsaws in a trending one. My eye accounts for that because I'm reading the larger context — what the higher timeframe is doing, what news is on deck, whether the move I'm watching is the third one of the morning or the first.

None of that lives in OHLC values. It lives in the rest of my screen.

What the experiment actually showed

The indicator is not useless. It's a great filter. When I trade my setup with the indicator's green/red zone as a tiebreaker, I take fewer marginal trades and the average quality goes up. The indicator was a bad replacement and a good assistant.

The lesson generalizes: tools complement, they don't replace. Anyone selling you a fully mechanical retail strategy is selling you the marketing, not the system.