How we test
Every catalog audit runs the same documented protocol. The protocol is part of the product, and deviations are printed in the audit.
| EA | free Market demo, the version installed on the test date (recorded in every audit; the vendor may already list a newer one) |
|---|---|
| Inputs | vendor defaults, apart from the lot inputs named below, inside one measurement frame for all runs (fixed base lot, 100,000 USD at 1:5000; the v1.3 rows below). We run the EA exactly as it installs and follow the vendor's published recommendations for symbol and timeframe. No tuning, no set-file shopping. Any deviation is recorded in the audit. |
| Symbol | the vendor's advertised main symbol |
| Chart timeframe | vendor recommendation, otherwise M1 (recorded) |
| Window | the full available tick history per symbol → test date (gold usually from 2003; a shorter window is printed in its audit), identical rule for all, so audits stay comparable across the catalog. v1.1/v1.2 audits used a fixed 2020-01-01 start. The window of every run is printed in its audit. |
| Data world | one price history per instrument: native IC Markets Raw Spread ticks from July 2023 (1 July for BTCUSD, 3 July for the others), and before that Dukascopy tick data re-based to a constant modelled spread. For gold (XAUUSD) that spread is 9 cents (IC quotes gold in a stable cent band, while the raw Dukascopy spreads are several times wider and would execute-kill any scalper artificially). The break point and the model character of the pre-2023 segment are stated in each audit. |
| Model | Every tick based on real ticks |
| Execution delay | 10 ms |
| Lots (v1.3) | 0.01 fixed base lot (or the EA's smallest fixed setting). Built-in progressions such as martingale or grid chains stay ON: they are the mechanics we measure. We switch off balance coupling where the EA exposes it as an input (percent risk, auto lot), so the dollar trade sequence becomes capital-independent. Where an EA still scales its lots on its own, the audit says so in its scope box and gates the survival map. Any changed input is recorded in the audit. |
| Deposit / leverage (v1.3) | 100,000 USD at 1:5000, identical for all runs. Both are measurement settings, not a recommendation. The deep buffer and maximal leverage keep the tester from censoring the trade sequence with margin rejections in almost every run; where an EA still uses up the frame or stops trading long before the window ends, the scope box of its audit gives the date. What a real account survives is then computed, not assumed: the death chain, the capital ladder and the what-if tables run on the vendor's recommended deposit (1,000 USD when none is published), and the percent return and drawdown are quoted arithmetically on that vendor deposit; CAGR, withdrawal percentages, the prop rule replay, the challenge simulation and the Monte Carlo drawdowns and ruin probabilities stay on the 100,000 USD run capital. |
Protocol versions. v1.1/v1.2 audits (the original catalog) used the vendor's recommended deposit and leverage directly. v1.3 (adopted 2026-08-16) introduced the standard measurement frame above. The two are not mixed inside one audit.
Trading costs in our data world. Catalog reruns execute on price histories built from Dukascopy and IC Markets tick data (see the data world above), one per instrument. Swap is modeled natively by the tester from each symbol's properties (for gold: -59.285 points long, +40.792 short, triple swap on Wednesday). We verified real swap deals in the reports. Commission is not modeled by the tester on our price histories, so the engine retrofits it deterministically at 3.50 USD per lot per side on every deal, rounded to the cent per deal like a broker statement (a 0.01 lot deal therefore pays 0.04 USD, not 0.035), rebuilds the balance chain, and prints the retrofit line in the cost waterfall. An EA that never holds overnight legitimately shows zero swap. The audit distinguishes that from a missing cost model.
The audit engine then reconstructs every trade from the deal list, validates the pairing against the tester log (or the report's own holding-time figures), verifies the balance to the cent, and runs the detectors and prop-firm simulations. Same inputs → same audit, bit for bit.
Run the same test yourself
- In MT5, open the EA's Market page and press Free demo. It downloads into your terminal for tester use.
- Strategy tester: choose the EA, symbol, M1 chart (or vendor recommendation), the custom period printed in the audit you want to reproduce, model "Every tick based on real ticks", 10 ms delay. To reproduce a v1.3 catalog run, set the deposit to 100,000 USD, the highest leverage your tester offers, and the EA's lot input to a fixed 0.01 base lot. For v1.1/v1.2 audits use the vendor's recommended deposit and leverage as printed in the audit.
- Before the run, clear the tester journal (Tester → Journal → right-click → Clear log). Do not try to delete the log files in Explorer, because MetaTrader keeps the current one locked. Old runs in the same journal blur the pairing.
- After the run, right-click the results → Report → HTML, and
save the tester journal/log file too. Right-click the Journal →
Open folder shows where those logs live. That is the path you
need when you pick the files for the upload. A run that crosses
midnight writes one file per date (
20260811.log,20260812.log, …). Keep all of them, they are one run. - Bring everything to Audit my EA.