Forex Trading Guidelines › Guideline 18
EA Trading Guidelines: 12 Rules for Automated Forex Robots
The 12 EA Guidelines
| # | EA rule | Why it matters |
|---|---|---|
| 1 | Hard stop loss on every position | No SL = one bad candle can blow the account. The EA must refuse to trade without a stop. |
| 2 | Risk-based sizing (0.5-2%) | Lots must be computed from equity and stop distance — never fixed lots that become oversized as equity falls. |
| 3 | Daily loss shutdown | Close all positions and stop trading for the day at a 3-5% daily loss (the 3-5-7 rule, automated). |
| 4 | Max consecutive losses pause | After N losing trades in a row, pause until the next day — markets regimes can flip a losing streak into a blow-up. |
| 5 | No martingale / grid / averaging | Position doubling is a countdown to zero. If the EA "needs" it, the edge does not exist. |
| 6 | News blackout filter | No new positions 30 min before/after high-impact events (Guideline 9). Spreads explode at releases. |
| 7 | Max spread check | Skip the trade if spread exceeds a multiple of the average — slippage and spread widen kill scalping edges. |
| 8 | Session filter | Trade only configured sessions; skip rollover and Friday close windows (Guideline 15). |
| 9 | Max open positions & correlated exposure | Cap simultaneous positions and combined exposure (3% open risk) — EURUSD and GBPUSD can act as one trade. |
| 10 | No Friday / expiry-day holding | Flatten before weekends and expiry when gaps can gap through every stop. |
| 11 | Forward-test on demo first | Minimum 1-3 months of demo forward-testing after backtest — the 90% rule (Guideline 6) applies to robots too. |
| 12 | Monitor daily; kill switch | A robot is not "set and forget" — review the journal daily and keep a manual kill switch for regime changes. |
The Backtest Red Flags
Before trusting any backtest result, check for these classic overfits — they are the reason most backtests fail live:
- Profit factor above 2.5 over years of data — suspiciously good; retail-quality data + realistic spreads rarely produce this.
- Zero or tiny slippage assumptions — real execution costs 0.5-2 pips more than backtests show.
- Curve-fitted parameters — a strategy that only works with 200 EMA and 14 RSI tuned to one historical window.
- Martingale or grid included — the backtest will show a beautiful curve right up to the vertical drop.
- Small sample — fewer than 300 trades, or fewer than 3 different market regimes, proves nothing.
Where to Encode These Rules
Use the Free MT4 EA Builder or Free MT5 EA Builder to turn a 9-part strategy (Guideline 11) into an EA, then verify the EA follows all 12 rules above before funding it. For fully custom robots, MT5 EA development services can encode every safeguard.
FAQ
Why do most forex EAs fail?
Most retail EAs fail because they lack risk controls (no stops, fixed oversized lots, martingale), were overfitted to a single backtest window, or were forward-tested for days instead of months. The 12 EA guidelines fix all three.
Is martingale EA trading safe?
No. Martingale doubles size after every loss — the backtest curve looks amazing until the streak that empties the account. It violates guideline 5 above and guideline 14 of the human rulebook.
How long should I test an EA on demo before live?
At least 1-3 months of forward testing across different market conditions, after a backtest on at least 300 trades covering multiple regimes. Then fund small and monitor daily.
What risk settings should my EA use?
0.5-2% risk per trade computed from equity, 3-5% daily loss shutdown, max 3% open exposure, stop loss on every trade, and no size increase after losses. These match the human guidelines and prop-firm rules.
Choose the next useful step
Most visitors arrive with one question: build, test, price, or trust. These shortcuts keep the path practical.