Our review methodology
Every grade on AlgoVerdict comes from six weighted criteria. The weights are listed here in full — they are read from the same configuration file that computes the grades, so they cannot drift from what the review pages show. Exception: our own products ↓
The weights
EA reviews
| Live track record | 25% |
|---|---|
| Strategy risk | 20% |
| Transparency | 20% |
| Value for money | 15% |
| Usability | 10% |
| Broker independence | 10% |
Broker reviews
| Costs | 20% |
|---|---|
| Execution | 20% |
| Algo suitability | 20% |
| Regulation | 15% |
| Conditions | 15% |
| Deposits & support | 10% |
Prop firm reviews
| Payout reliability | 25% |
|---|---|
| Challenge cost | 20% |
| Profit split | 20% |
| Rules | 15% |
| EAs allowed | 10% |
| Scaling | 10% |
Where the live figures come from
For EAs the live track record carries the most weight — so we distinguish how solid a piece of evidence is. Marketing backtests count in no class.
- Class A — run by us or ATS: a real account whose figures are published. The mark follows the account age, not the return.
- Class B — platform-verified vendor record: a public MQL5 signal or a Myfxbook account flagged REAL. It counts, but as second-class evidence.
- Class C — unusable evidence: demo account, a signal switched off or swapped out, or the signal of a different product. Earns nothing.
- Class D — no evidence: we say so explicitly on the review page.
The fingerprint
Our own accounts often run more than one EA, so an account loss is not automatically the loss of one particular EA. We separate the tracks by each trade's magic number and order comment, and measure the decline a single EA caused on its own. That figure adds to the published accounts, it does not replace them: the grade always follows the worst figure from any source.
The traffic light
- A grade below 3.0, or a measured decline of 50% or more → red.
- Above 50% there is exactly one exception: if a calm, long-running account proves a calibrated way to run it exists, the light is yellow rather than red.
- From 30% decline upwards, green is out.
- Green from a grade of 3.8, otherwise yellow.
What we cannot do
We cannot predict future returns and do not try. A long, clean track says a strategy has survived several market regimes — not that it will survive the next one. With grid and martingale systems the heavy loss is rare and arrives all at once; a quiet year proves nothing there. And where data is missing we rate conservatively instead of filling the gap with vendor claims.
Independence
No vendor pays for a rating or sees one before publication. We are partly funded by affiliate links to brokers and VPS providers; that is marked wherever it appears and changes neither grades nor traffic lights. Prices are taken only from vendor, ATS or MQL5 product pages, never from marketing copy — that often quotes different figures than the buy button.
We do not rate our own products
The operator of AlgoVerdict develops and sells his own Expert Advisors and MT5 tools, such as Dawnbreak, LongRunner or Drawdown Minimizer. These products are not rated: no stars, no grade, no traffic light, no ranking and no figures from our own live accounts. The reason is simple: whoever earns from a sale cannot rate his own product independently – not even with the same weights as for other vendors’ EAs, because the choice of evidence and every judgement call would still be ours. Until 22 Oct 2026 we graded Dawnbreak with the same methodology; since then this rule applies to every product of our own.
What stays on these pages: verifiable facts with source and date, the code we read, the developer’s backtests – labelled as backtests –, prices, strengths and weaknesses, and the link to public signals where one exists. Each page opens with the disclosure that we earn from sales; the product link is labelled “our own product”, not “partner link”.