Testing & evaluation
My AI Gold MT5 1.71: one year on XAUUSD
From 2025-10-12 to 2026-10-12, the fixed-size test finished at +US$1,969.69 after recorded costs, with maximum percentage equity drawdown of 11.46%. This review examines the setup, weaker periods and actual simulated trades behind the result.
Tested and written by MQLume, the product seller. This is a historical simulation, not a live account track record or an independent endorsement.
The fixed setup
This XAUUSD EA exposes controls for entries, trading sessions, stops and trailing management. Lot is 0.01 and recovery is disabled. Other strategy settings retain their delivered defaults.
Inputs were frozen before reviewing annual performance. This run was not optimized, and the delivered executable was tested without strategy-code changes. Fixed 0.01 lot refers to each executed trade deal, not a fixed percentage of account risk.
| Data source | IC Markets / ICMarketsSC-MT5-6 |
|---|---|
| Instrument / chart | XAUUSD · H1 |
| Period | 2025-10-12 — 2026-10-12; end exclusive, broker time |
| Simulated capital | US$1,000 · leverage 1:100 · hedging |
| Position size | Fixed 0.01 lot |
| Execution model | Every tick based on real ticks · 150 ms delay |
| MT5 history quality | 99% |
Profit and drawdown together
The net result was +196.97% of the initial US$1,000. MT5 reported a profit factor of 2.59 and 200 (83.68%) winning trades. Maximum balance drawdown was US$110.77 (8.04%). The largest equity decline in dollars was US$158.13 (11.46%); the largest percentage decline was 11.46% (US$158.13). These maxima can occur at different points. Equity includes floating profit and loss, so balance alone does not describe all account risk.
The largest combined open size in the deal record was 0.02 lots. The longest observed wait for a new balance high was 44.73 calendar days, measured at deal times. This is not an intraday equity recovery measure. Simultaneous positions, stop distances and market gaps all affect risk.
The monthly pattern
The strongest calendar segment was 2026-08, at +US$275.39; the weakest was 2026-04, at +US$57.41. The first and final October segments are partial months. All segments, including losses and months without trades, remain in chronological order.
Costs included in the result
Trading profit before separately charged fees was +US$2,009.86. Commission was −US$19.12 and swap −US$21.05, reconciling to net +US$1,969.69. Bid/ask spread is already reflected in execution prices and has not been subtracted again. Broker contract settings and costs can change; these figures do not establish future execution conditions.
Inspect actual trade sequences
The selection rule is the first profitable trade, the first losing trade, and the largest paired loss when it is different. Net figures include both entry and exit commission and recorded swap. They may differ from platform statistics that allocate entry fees separately.
| Selection | Entry / exit time | Direction / prices | Net P/L |
|---|---|---|---|
| First profitable trade | 2025.10.13 09:00:00 2025.10.13 12:02:35 | Buy 0.01 4,063.44 → 4,079.38 | +US$15.86 |
| First losing trade | 2025.10.17 07:00:00 2025.10.17 10:20:26 | Buy 0.01 4,366.90 → 4,326.09 | −US$40.89 |
| Largest paired trade loss | 2026.04.22 15:00:00 2026.04.23 03:11:52 | Buy 0.01 4,750.90 → 4,710.46 | −US$42.23 |
What the test cannot establish
MT5 lists History Quality at 99% in real-tick mode. This is a data-quality indicator, not a prediction of live accuracy. The tester can generate ticks from minute bars when real ticks are missing; selecting the mode alone does not prove complete coverage. The fixed delay cannot reproduce every liquidity condition, execution issue or operational failure. Pending-order execution follows tester rules.
This is one broker feed, one period and one predeclared configuration. It does not establish optimal settings or predict the next year. No demo forward-test result is included. Reproduce a baseline, inspect losses and test on a demo account before considering live use.
Make the source your own
Editable source lets you investigate an idea such as a trading-session filter or clearer logs. Work on a separate copy, change one thing at a time and compare with the unchanged baseline. An AI coding assistant can explain code and draft changes; you still need to review, compile and test them.
Trading involves risk. Past and simulated results do not guarantee future returns. This source product is intended for users with programming and trading knowledge.