Ayaz

MT4 and MT5 Backtest Auditor

Freelance • Indépendant

Feb 2024 - Present • 2 yrs 7 mos

Testing Expert Advisors on gold and other volatile symbols, mostly long-horizon runs over ten to sixteen years of history. The work starts with the data, because that is where most backtests quietly fail. Real tick coverage is checked month by month, so it is clear which part of a test ran on the broker's actual ticks and which part the tester generated from M1 bars. On one gold symbol I found sixteen consecutive months with M1 bars and zero real ticks, all of it still producing a clean-looking result. The symbol specification is then read from the terminal rather than assumed: contract size, digits, point and tick value. Two symbols on the same broker, names differing by one letter, can carry contract sizes of 1 and 100, which makes the same lot size a hundred times the risk with nothing in the report to show it. Drawdown is recalculated from peak equity instead of balance, because a balance figure goes blind the moment an account is holding losing positions open. Grid and martingale baskets are measured for how long they stay open and how far they run against the account while the equity curve still looks calm. Costs are modelled explicitly, including commission expressed in spread points so it scales correctly at every lot size. Each test is re-run on the real-tick period alone, and a walk-forward is run on a stretch the settings were never tuned on. Where a live account exists, the live trade history is reconciled against the tested result. Source reviews cover the risk traps a backtest never shows: contract size hardcoded instead of read from the symbol, drawdown measured against balance, position sizing with no ceiling, weekend and Friday exposure. Every report states where each number came from, so a client can hand it to the vendor and invite the recheck. What the work does not do is predict whether an EA will make money.