What Auto-Trading Really Is: From Rules to a Testable MT5 EA

1. What automation actually changes

An Expert Advisor can execute predefined rules and controls in MetaTrader. Running code is not evidence of a trading edge. AI may help explain a report or draft code, but its output still needs verification. This lesson produces a testable system description, not a ready-made profitable robot.

2. Separate four responsibilities

Signals describe observed conditions. Risk controls decide whether trading is permitted and bound the exposure. Execution sends orders and handles broker responses. Logging records decisions, versions and settings. A rejected request or a disconnected terminal should not cause unlimited duplicate orders.

3. Make the rule testable before coding

“Buy when the trend is good” is ambiguous. An educational specification can instead say: evaluate only after a bar closes, record one decision per bar, and block new entries when data or connection state is invalid. Trend definition, exits, size and signal expiry still need precise definitions. This illustrates specification quality; it is not a recommended trading strategy.

4. Use an honest validation sequence

Keep a baseline with a frozen build, saved configuration, symbol, period and execution assumptions. Include commission and spread, and document slippage assumptions. Reserve out-of-sample data that was not used to select settings. Repeatedly adjusting parameters after observing that segment turns it into development data; do not continue describing it as untouched final evidence.

5. Assignment and acceptance criteria

Complete these fields: symbol, timeframe, entry, exit, stop logic, sizing, daily loss cap, maximum concurrent positions, no-trade conditions and reconnect behaviour. Write an example input and expected output for each. Then plan historical, cost-stress and demo-forward tests. Report drawdown and sample size beside profit. Successful tests do not guarantee future returns.

Assignment and next step

Write a ten-line EA specification before writing any MQL5.

Lesson 2: Strategy Tester metrics and evidence.