EA reviews · Mike Harty · Published 2026-10-24 · Updated 2026-10-24

How We Review EAs: The VRD EA Risk Score Methodology

Short answer: We score every expert advisor from 0 to 10 on six risk dimensions: drawdown, grid/martingale exposure, capital needed, consistency, quality of evidence and vendor transparency. We weight them, then cap the total when there are hard red flags. Higher means lower, better-evidenced risk. It is not a profit score.

Key takeaways

• We rate risk and evidence, not how much an EA might make.
• Live, verified, deposit-clean signals count far more than backtests.
• Grid, martingale or averaging without a hard stop caps the score, whatever the returns.
• Anything we can't verify is marked Unverified, and it scores as if absent.

What our scores are based on (read this first)

We do not currently hold licensed copies of the EAs we review. Until we do, every score is based on only two kinds of evidence: 1. Public signal data: MQL5 signal pages and other public, broker-linked track records, plus the vendor's own public product pages and documentation. 2. Backtests we have run ourselves in the MT5 Strategy Tester, where possible. Each review lists the exact settings, period, broker data and history quality.

What this means in practice: - We have not run EAs on live or demo accounts of our own. No review contains VRD live results. - Backtests are historical and hypothetical. We label them clearly and never present them as live results. - Monte Carlo and slippage stress tests are run on backtests only (method: Monte Carlo for MT5 backtests). A high blow % means the worst single event is bigger than the floor. It is not a forecast of how often the account will blow. We have not run execution-delay tests. - If a review has no backtest of ours, it is scored on public evidence alone, and the review says so.

1. The six dimensions and weights

VRD Risk Score = Σ(score × weight), rounded to 0.1, then red-flag caps applied (section 3).

2. Scoring scales (0–10)

D1 Drawdown (25%)

We use the maximum equity drawdown (includes floating losses) from the longest verified live signal. If only balance drawdown is published, we treat it as a lower bound and drop one band. Relative DD (% of peak) beats absolute DD ($).

Adjustments: −2 if equity DD is far above balance DD (more than 2×, a sign of hidden floating losses). −2 if deposits were made during a drawdown (the DD is understated).

D2 Grid / martingale / averaging exposure (20%)

Evidence for D2: the trade history (look for clustered opens, increasing lots, shared TP), the strategy description, the inputs (e.g. "LotMultiplier", "GridStep"), and independent analyses.

D3 Capital needed (15%)

Compute the realistic deposit, the one that survives a 1.5× worst observed drawdown at the vendor's default settings, and compare it with the vendor's minimum (see the capital cornerstone for the method).

Override: if the vendor's own suggested deposit at the tested lot is below the worst backtest equity drawdown, D3 is capped at 4.

D4 Consistency (15%)

Monthly returns on the verified signal: % of positive months, worst month, and whether returns depend on a few outlier months. If the vendor runs several signals, do they agree? 10 = ≥18 months, ≥75% positive, worst month > −5%; 5 = 6–12 months or mixed; 0 = < 3 months or one month makes most of the gains.

D5 Evidence quality (15%)

Backtests alone can never score above 1, because they are curve-fittable and usually ignore latency, slippage and variable spreads.

Our own backtests: a multi-year backtest we ran ourselves (at least 5 years, 100% history quality, settings published) adds +1 to D5, up to a maximum of 7. It never replaces a real-money signal. Where the vendor's recommended lot is larger than 0.01, we test at that lot too, because recovery and multi-leg EAs don't scale linearly.

D6 Vendor transparency (10%)

Points (out of 10): strategy honestly described, including grid/martingale (3). Real-money signal linked from the product page (2). No history of delisted or relaunched products (2). Changelog and settings documentation (1). Responsive to negative reviews without deleting them (1). No profit promises in the copy (1).

3. Red flags: hard caps

The lowest applicable cap wins.

4. Evidence rules

1. Every number has a source link and the date we accessed it. Screenshots get archived (Wayback or local) because MQL5 pages change.
2. Unverified = absent. A vendor claim we can't check is labelled UNVERIFIED and scored as missing evidence.
3. Primary first: MQL5 signal page > broker statement (Myfxbook/FXBlue with verified track record) > product page > third-party analysis > user reviews.
4. Third-party scores (ForexInspector, AlgoCheck, etc.) are cited as their opinion, never copied in as our measurement.
5. Right of reply: we contact the vendor before publishing and include their response.
6. Re-review every 6 months, or after any major version or delisting. The page shows "Last reviewed" plus a changelog.
7. No conflicts hidden: we sell MT5 risk tools, not EAs. We take no affiliate fees from EA vendors. If that ever changes, we disclose it on the review.
8. Every review is fact-checked by a second person before publishing.

5. Review page structure

Short answer (score + one-line verdict) → Key facts table (price, platform, symbol, strategy type, signals, sources) → D1–D6 sections with evidence → Red flags → Who it might suit / who should avoid → Capital estimate → FAQ → Sources list → Changelog → Disclaimer.

Risk disclaimer: Trading forex and CFDs with leverage carries a high risk of loss. Reviews describe historical, publicly available evidence and our own backtests on the date stated. Backtests are hypothetical, not live results. They are not investment advice and they don't predict future results. Vendors may change products at any time.

Frequently asked questions