How We Test Forex EAs

Testing an EA properly takes time, discipline, and the right infrastructure. Most review sites skip the hard part — they run a quick backtest, glance at the vendor's marketing page, and write up a glowing recommendation. We don't work that way.

Every EA we review goes through a multi-stage process that includes forward testing on live demo accounts, backtesting with quality tick data, risk analysis, and a detailed evaluation against our scoring criteria. This page explains exactly what that process looks like, so you know what's behind every score we publish.

Our Testing Environment

Running an EA on your home PC with a spotty internet connection doesn't give you reliable data. That's why we run all our tests on dedicated VPS servers designed for forex trading.

Infrastructure Specs

VPS Provider: ForexVPS / BeeksFX
Location: New York / London
Platforms: MT4 & MT5
Latency: <5ms to broker
Primary Brokers: IC Markets, Pepperstone
Account Type: Demo (Raw Spread)
Spread Conditions: Variable, from 0.0 pips + commission
Uptime: 99.9%+

We use demo accounts with raw spread conditions to match what most retail traders would experience on a live account. The VPS ensures consistent execution and eliminates connection drops that could skew results. If an EA is sensitive to spread or latency, we note that in the review.

Forward Testing

Forward testing is the backbone of our review process. Backtests can be curve-fitted. Vendor screenshots can be faked. But a live demo account running for months in real market conditions? That's much harder to fake.

Here's what our forward testing process looks like:

• Minimum 3 months of live demo testing. We don't rush reviews. Three months gives us enough data to see how an EA handles different market conditions — trending, ranging, volatile, and quiet periods.
• Verified Myfxbook tracking. Every forward test is connected to a Myfxbook account with verified trading privileges. We link to these in our reviews so you can check the numbers yourself.
• Real-time slippage and execution data. We track actual execution prices against requested prices. Slippage matters — especially for scalpers — and we report it honestly.
• Default settings first. We initially run every EA on its recommended default settings. If results are poor, we may optimize — but we always report what happened out of the box, because that's what most buyers will experience.

We've published reviews where an EA looked great for 6 weeks, then gave back all its gains in week 7. That's exactly why the minimum test period matters. Short tests hide drawdowns.

Backtesting Standards

We don't rely on backtests as our primary metric — forward testing always takes priority. But backtests are useful for understanding an EA's behavior across longer time periods and market events that a 3-month forward test won't cover. The key is doing them right.

Our Backtesting Parameters

Tick Data Quality: 99.9% modeling quality using Tick Data Suite or Dukascopy tick data. MT4's built-in data at 90% quality is unreliable for any serious analysis.
Test Period: Minimum 10 years (2015-2025). We need to see performance through multiple market regimes including COVID volatility (2020), the 2022 rate hike cycle, and recent conditions.
Spread Simulation: Variable spreads based on historical data, not fixed. Fixed spread backtests are meaningless for EAs that trade during high-volatility sessions.
Monte Carlo Analysis: We run Monte Carlo simulations to stress-test results. This randomizes trade order and applies worst-case scenarios to see how an EA might perform under different conditions.
Walk-Forward Analysis: When possible, we perform walk-forward optimization to check whether an EA's parameters are curve-fitted to historical data or have genuine predictive value.

If an EA's backtest looks stellar but its forward test is mediocre, that's a red flag for curve-fitting. We always note when there's a significant gap between backtest and live performance.

Scoring Criteria

Every EA receives a score out of 10, calculated from five weighted categories. Here's exactly what each one measures and how it's weighted:

Performance

30% weight

The bottom line: does this EA make money? We measure absolute returns, risk-adjusted returns (Sharpe ratio), maximum drawdown, and recovery factor. An EA that returns 50% with 10% drawdown scores higher than one returning 100% with 60% drawdown. Consistency matters more than peak numbers.

Risk Management

25% weight

How does the EA protect your capital? We evaluate lot sizing logic (fixed, percentage-based, or martingale), stop loss implementation, maximum drawdown controls, and correlation management when trading multiple pairs. EAs that use martingale or grid without stop losses get penalized heavily here.

Transparency

20% weight

Does the vendor provide verified live trading results? Is the strategy logic explained, or is it a complete black box? Are there honest risk disclosures, or just hype about returns? We check for verified Myfxbook/FXBlue accounts, transparent refund policies, and honest marketing. Vendors who doctor results or make unrealistic claims score poorly.

Value for Money

15% weight

Is the EA worth what the vendor charges? We consider the pricing model (one-time vs. subscription), number of accounts included, features relative to price, and whether free or cheaper alternatives deliver similar results. A $500 EA needs to justify that price with performance a $100 alternative can't match.

Support & Documentation

10% weight

Does the vendor provide clear setup guides, responsive support, and regular updates? We test support response times, check documentation quality, and evaluate whether a beginner could set up the EA without pulling their hair out. Vendors who ghost customers after the sale lose points here.

What We Check For

Beyond our scoring criteria, we specifically investigate common tricks and dangerous strategies that vendors often hide in their marketing:

Hidden Martingale

Some EAs advertise "smart money management" but actually double lot sizes after losses. We decompile settings files and monitor lot sizing in live tests to detect this. Martingale will blow your account eventually — it's a question of when, not if.

Grid Without Stops

Grid strategies can look profitable for months until a strong trend wipes out the account. We check whether grid EAs include hard stop losses and maximum drawdown limits. If they don't, we flag it.

Curve-Fitted Backtests

An EA optimized to historical data can show 1000% returns in backtesting but fail immediately on live markets. Walk-forward analysis and out-of-sample testing help us detect this.

Fake Myfxbook Results

Rented accounts, demo-only results presented as live, and accounts that conveniently go private after drawdowns. We verify account ownership, check for live/demo status, and document the common tricks vendors use to fake results.

Our Review Policy

A review isn't a one-and-done document. Markets change, EAs get updated, and performance shifts over time. Our policy reflects that reality:

• We update reviews when EAs update. If a vendor releases a major version update, we re-test and revise our review accordingly. The date at the top of each review reflects the last update.
• Quarterly re-evaluation. Every EA in our active testing pool gets reviewed quarterly. If performance has degraded significantly, we adjust the score downward and explain why.
• Score drops happen. An EA that scored a 7 last quarter might score a 6 this quarter if forward test results have deteriorated. We don't let old scores linger when they no longer reflect reality.
• Discontinued or scam EAs get flagged. If a vendor disappears or we determine a product is a scam, we update the review with a clear warning and remove any affiliate links.

Our testing methodology isn't perfect — no methodology is. Demo accounts don't perfectly replicate live execution, and 3 months of forward testing won't catch every market regime. But it's a far higher bar than what most of this industry demands, and we think traders deserve at least that much diligence before being told to risk their money.