{
  "id": 11489781,
  "title": "Your EA crashed at 3am — who hits the brake?",
  "url": "https://urgent.news/2026/10/02/your-ea-crashed-at-3am-who-hits-the-brake",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-10-02T17:22:11.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/xuks124/your-ea-crashed-at-3am-who-hits-the-brake-5caj"
  },
  "original_language": "en",
  "account": "At 3:07am, your virtual private server (VPS) has been experiencing issues since midnight. An Expert Advisor (EA) that automatically opens breakout trades on M5 bars experienced a trade order error around 30 minutes ago, and continued executing trades without any intervention. When you wake up at 7am, you find that there are now eleven unintended positions, a significant account drawdown you must explain to yourself, and a lingering question of who is responsible for shutting down the EA.\n\nMost EA setups have no mechanism to stop trading, relying solely on take-profit and stop-loss orders. However, experts who have used EAs for a long time are aware of the limitations of these orders, as they do not cover all possible failure modes. These include the terminal or connection dying mid-trade, the stop-loss not functioning correctly, or the position that was intended to be closed at the worst possible moment never getting the response. The EA continues to execute trades based on a strategy that has become irrelevant.\n\nErrors are logged but rarely acted upon, and MQL5 error codes scroll through the logs while the process remains technically running. A log file is not sufficient supervision. You are asleep, and the entire reason for automating the EA is for risk management. The system's built-in risk tools on the order side (stop-loss, position sizing) assume that trades will happen as planned. However, the 3am problem is different: the system misbehaves due to a crash, freeze, runaway loop, or disconnected feed, and someone must notice and act.\n\nA brake layer is not a strategy and does not need to be clever. It must be straightforward and awake, monitoring the process rather than the profit and loss (P&L). It should detect abnormal states such as a frozen chart, an ever-growing position count, error bursts, or a terminal that restarted without the expected state. These are anomaly checks, not predictions – nobody needs to forecast prices to notice that the engine is on fire. The human should have a switch to stop opening new trades, close all existing positions, and flatten them immediately.\n\nThe value of a manual kill switch is that it works even when you are half-awake and scared, without the need for configuration or terminal adjustments. The watchdog should fail towards safety, and if it loses sight of the terminal, the conservative default should be to stop trading rather than continue and hope. A brake layer cannot make an EA profitable, but it can cap the tail – the nights when a bug rather than a market decides your drawdown. Prioritizing survival first and edge second is essential. Remember that a brake layer cannot make an EA profitable; it can only limit the worst-case scenario.",
  "summary": "It is 3:07am. Your VPS has been flaky since midnight. The EA that opens breakout trades on M5 bars threw a trade order error thirty minutes ago — and kept going, because nobody told it to stop. When you wake up at 7am, there are eleven positions you never intended, an account drawdown you have to explain to yourself, and one question you keep postponing: who hits the brake? The honest answer:…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}