MQL5 Market / Fil d'actualité

A clean recompile is where an MT4-to-MT5 migration starts, not where it ends. The functions rename, the errors clear, the Strategy Tester runs — and four architecture differences sit underneath, none of which throws an error...

If you are migrating from MQL4 or debugging a position management bug you cannot reproduce in Strategy Tester, read this before changing anything else. What the Documentation Actually Says The MQL5 reference for CTrade::PositionClose() is technically accurate. Both overloads are listed...

There is a conversation that repeats often enough to be worth writing down. A trader has traded a setup manually for months, decides to automate it, and frames the project as "turn my rules into an EA." The entry logic gets written in an afternoon...

Every EA project starts with a brief. The brief describes entry conditions, exit logic, position sizing, and timeframe. It is usually thorough on paper. In practice, it is almost never complete...

The Failure Pattern Most EA reliability discussions focus on broker edge cases, slippage, and backtest-to-live gaps. The restart failure is quieter — and in my experience, more consistently expensive. The pattern: an EA runs correctly for months on a VPS...

The Problem Three EA rescue projects arrived at barmenteros FX this year with identical symptoms: excellent single-trade backtest results, inconsistent live performance during periods when two or more positions ran simultaneously...

Permission to run an Expert Advisor is not what disqualifies automated prop-firm accounts. Automation permission is not where firms compete — they differentiate on payout splits, drawdown limits, and challenge fees...

The Silent Failure Mode Running two or more expert advisors on the same MetaTrader account exposes a structural gap that single-EA backtesting never reveals. The platform logs no warning when one EA modifies or closes an order it doesn't own...

A PPO agent trained for two million steps on EURUSD 15-minute data with flat reward curves across all five seeds is not an optimizer problem. Switching to SAC will not fix it. Changing the learning rate will not fix it...

Most DRL algorithm comparisons you find in the literature rank models by annualized return. If you plan to deploy the agent on a live MetaTrader account, that is the wrong column to sort on. Here is a benchmark that makes the point concrete...

If you have trained a deep reinforcement learning agent in Python and tried to run it against a live MT5 terminal, you may have seen this: the agent converges cleanly, the backtest Sharpe looks good, and then on a paper account it behaves like it never trained. No error, no warning...

A client asked me to diagnose why his FTMO Phase 1 EA was disqualified on day 9. The strategy was profitable — positive P&L at the point of disqualification. The journal told the story: trades were opening after the daily loss limit had already been breached...

A clean compile tells you the syntax is valid. That is the whole of its claim. It does not tell you that the correct order was selected before the close, that a rejected request was noticed, or that the Expert Advisor still knows what it holds after a terminal restart...

The Pattern A trader runs an EA profitably for months or years. No code changes. No settings modified. The equity curve starts bleeding. The trader backtests — still shows good results. The trader re-optimizes — makes it worse. I have encountered this three times since January 2026...

The client who ran double risk for six weeks Last year I reviewed a client's XAUUSD Expert Advisor. The risk input read 2% per trade. The actual exposure per position was 4.1%. Every trade for six weeks had been roughly double the intended size...

I review AI-generated MQL code regularly in rescue projects. Since 2024, the same three failure patterns appear in the majority of AI-generated EAs. The code compiles clean and backtests well. It fails in production because it misses the conditions that only exist when real money is on the line...