Methodology
This is how we review an EA. A score is evidence, not a claim the robot is legit.
Data
Real tick data. Variable spread, commission, 100ms slippage model. No M1 interpolation.
Grades
| A+ | ≥90 |
|---|---|
| A | ≥85 |
| A- | ≥80 |
| B+ | ≥75 |
| B | ≥70 |
| B- | ≥65 |
| C+ | ≥60 |
| C | ≥55 |
| C- | ≥50 |
| D | ≥40 |
| F | <40 |
Dial zones: <40 Risk, 40–70 Amber, ≥70 Signal.
Lab replay
Catalogue Market Free Demo builds can be replayed on an IC Markets demo with default inputs, one symbol, and a 100 ms execution delay (not spread slippage). That replay is not a source audit and never grants the verified seal.
Lab score = round(window performance × 80% + source trust × 20%). The same checklist is used for every EA with a source audit. There is no separate per-slug formula. Window performance (replay-v3) is 40% capital preservation, 40% profit factor, and 20% share of profitable windows. Source trust is the equal-weight checklist already on the audit row: each recorded check is 100 (pass) or 50 (warning). An n/a row is omitted, not scored as a pass. Martingale, grid, and a hard stop must be recorded, and any failed check withholds the lift, so backtests alone cap at 80. The Verified seal is a separate certification and does not move with this number. Seller evidence is compared on recorded conditions and shown side by side; it never changes the score, and no similarity percentage is computed.
Limits
A backtest is evidence, not a promise. We publish what we cannot test: future regime change, broker execution drift.