The deliberately boring system
We're going to build a moving average crossover strategy. Yes, the most clichéd system in existence. And that's entirely intentional.
Why boring?
This series isn't about discovering alpha. It's about learning the complete development lifecycle of an automated trading system — from specification through implementation, backtesting, validation, and robustness analysis.
A simple system lets us focus on the process. If we started with something complex, every debugging session would become "is the logic wrong or is the infrastructure wrong?" With a moving average crossover, the logic is so transparent that any issue we encounter is almost certainly infrastructure, implementation, or methodology — exactly the things we need to learn.
The specification
Following the five-component framework from Lab Note #01, here is our system specification in plain English:
| Component | Specification |
|---|---|
| Instrument | EURUSD |
| Timeframe | H1 (1-hour bars) |
| Data observed | Close price of each completed bar |
| Indicators | 20-period Simple Moving Average (fast MA) 50-period Simple Moving Average (slow MA) |
| Entry — Long | Fast MA crosses above Slow MA on the most recently completed bar. Enter at market on the open of the next bar. |
| Entry — Short | Fast MA crosses below Slow MA on the most recently completed bar. Enter at market on the open of the next bar. |
| Exit | Reverse signal only. When a new entry signal fires in the opposite direction, close the current position and open the new one. |
| Position sizing | Fixed: 0.1 lots per trade |
| Stop loss | None (initial version — we'll address this) |
| Execution | Market orders, assume 1 pip slippage per side |
| Costs | Spread: use live/historical spread from broker. Commission: as per account. |
What this specification gives us
Notice what we've done: every decision is explicit. Another developer could implement this system from this table alone and get identical results. There's no ambiguity about:
- When we calculate the signal (on bar close, not mid-bar)
- When we execute (next bar open, not the bar that generated the signal)
- What kind of orders we use (market, not limit)
- Our slippage assumption (1 pip — probably generous for EURUSD H1)
What we're deliberately ignoring (for now)
This specification has obvious weaknesses. No stop loss. No position sizing based on risk. No consideration of trading sessions or news events. Fixed parameters. Single instrument.
Good. We know these are weaknesses. We'll address each one as the series progresses. Trying to build the "complete" system on day one is how projects stall — and how complexity hides bugs.
In the next lab note, we take this specification and turn it into working MQL5 code — a real Expert Advisor that can run in MetaTrader 5's Strategy Tester.
Want help writing a rigorous specification for your own strategy?
A clear spec is the foundation of everything that follows. Let's talk.