Trading System Design: From Idea to Rules

An idea is not a system. A system is an idea reduced to rules precise enough that a stranger could execute them.

What a trading system actually is

A trading system answers five questions for every position:

1. What to trade (universe, instruments)
2. When to enter (setup, trigger)
3. How much to risk (position size)
4. When to exit at a loss (stop)
5. When to exit at a profit (target or trail)

If any of these is left to "judgment," you don't have a system — you have a habit. Habits cannot be backtested, improved, or trusted under pressure.

Stage 1: The hypothesis

Every system starts as a hypothesis about a market edge. Examples:

• "Trend-following works because participants anchor to recent prices."
• "Volatility expands after quiet consolidation."
• "Liquidity rests above equal highs; price tends to sweep it."

Write the hypothesis in one sentence. If you can't, the idea isn't clear enough to become a system.

Stage 2: Formalize the universe

Define which instruments qualify:

• Asset class (forex, futures, stocks, crypto)
• Liquidity threshold (e.g., average daily volume > $100M)
• Volatility band (e.g., ATR > 0.5% of price)
• Session (e.g., London + New York overlap)
• Timeframe (e.g., H1 entries, daily trend filter)

A vague "I trade EURUSD" is not a universe. "I trade EURUSD on H1 during London and New York, only when the daily ATR exceeds 0.6%" is.

Stage 3: Entry rules

Convert the setup into mechanical triggers:

Component	Example
Trend filter	Daily close above 200 EMA
Setup	Pullback to 21 EMA
Trigger	Bullish engulfing on H1
Confirmation	RSI rising from oversold
Filter	No trades 30 min before/after high-impact news

Each component should be binary: yes or no. If you can argue about whether the trigger fired, it's not a rule.

Stage 4: Stop loss

Pick a stop method that matches the setup's logic:

• Structure-based: below the swing low of the entry
• Volatility-based: 1.5 × ATR
• Percentage-based: 1% from entry (least defensible — ignores structure)
• Time-based: exit if not profitable after N bars

Avoid "mental stops" — they don't survive drawdowns.

Stage 5: Take profit

Method	When to use
Fixed target	Range-bound setups with measured moves
Trailing stop	Trend-following setups
Scale-out	Hybrid: take 50% at 1R, trail the rest
Time exit	Fade setups that should resolve quickly

A defined exit rule prevents "I should have taken profit" hindsight bias from corrupting your journal.

Stage 6: Position sizing

Tie sizing to risk, not to lot counts:

Size = (Account × Risk%) ÷ (Entry − Stop)

Document the risk percentage for every setup type. Different setups can have different risk budgets.

Stage 7: Edge documentation

Before backtesting, write down why you expect the system to work:

• What behavioral bias is being exploited?
• What participant group is on the losing side?
• Under what market regime will it fail?

If you can't articulate the edge, the rules will fit noise rather than signal.

The complete specification template

Universe:    [instruments, session, timeframe]
Trend filter: [...]
Entry setup:  [...]
Entry trigger:[...]
Stop loss:    [method, distance]
Take profit:  [method, target]
Position size:[risk %, formula]
Filters:      [news, max trades/day, regime]

If another trader can read this template and execute the system identically, it's ready to backtest.

Common design errors

• Over-specification: 12 filters means no trades ever fire
• Under-specification: "buy at support" without defining support
• No exit rules: profit taking becomes emotional
• No regime filter: trend system run during chop bleeds slowly
• Curve-fit parameters: every constant tuned to past data

From spec to test

Once the rules are crisp, run them through historical data using the methodology in the next article. A system that survives backtesting and forward testing only earns the right to be traded live — and even then, with the smallest size.

---

Next: positive expectancy — the math that decides whether the system survives.