Building a first expert advisor: scope it small

[Shadow Trader]

The classic beginner plan: build a fully automatic system that trades ten pairs, manages risk dynamically, reads the news, adapts to regimes and prints money. The classic beginner outcome: a half-finished mess of code that does none of that.

A smarter first project is tiny. One pair. One entry rule. One exit rule. Fixed risk per trade. A hard daily loss limit. Logging for every decision. That's it.

Why so small? Because the hard part isn't the trading logic. It's everything around it: handling rejections, reconnections, rounding lot sizes, time zones, restarts, and the dozens of ways code can misbehave when it meets a real broker.

A simple system exposes those problems in a manageable form. Once it runs reliably for a few weeks on demo, extra features can be added one at a time, each tested before the next arrives.

Complexity can always be added later. Reliability has to be built first.

What did the first automated system look like, and which unexpected problem taught the most?