12 reasons trading bots die in the first month (and the fix for each)
Most retail trading bots stop working within weeks. The causes are boringly repeatable: alerts on every tick, no daily loss limit, a laptop that sleeps. Here are the twelve and the fix for each.
Ranjit Prasad · September 18, 2026
Ask anyone who has run a trading bot for more than a year and they will tell you the same thing: the strategy was rarely what killed it. The bot died because of a setting, a missing limit, or a server that went to sleep. The good news is that the causes repeat, which means the fixes can be written down in advance. Here are the twelve I see most, roughly in the order they tend to happen.
1. The alert fires on every tick. A condition that is true for the whole candle fires again and again, and the bot opens five positions from one signal. Fix: set every alert to "once per bar close" and make the bridge reject a second order in the same direction while one is open.
2. There is no daily loss limit. A strategy that loses eight trades in a row on a news day is normal; a bot that keeps trading through it is not. Fix: a daily loss limit that stops all trading until the next day, enforced in the bridge or MetaTrader, not in your head.
3. Position size is a fixed lot instead of a percentage. The account shrinks, the lot stays the same, and the risk per trade quietly doubles. Fix: risk a fixed percentage of the current balance (one percent is the usual starting point) and let the message or the bridge compute the lot from the stop distance.
4. The symbol does not match. TradingView says XAUUSD, the broker says GOLD or XAUUSD.m, and the order is rejected or, worse, placed on the wrong instrument. Fix: a written symbol map, checked on demo, for every market the bot trades.
5. The laptop went to sleep. Or the Wi-Fi dropped, or Windows updated. The alert fired at 3 a.m., nobody was listening, and the stop never got placed. Fix: run MetaTrader and the bridge on a small rented server (VPS) with automatic restart, and make the bot send you a heartbeat message so silence is noticed.
6. Automated trading was switched off after a restart. MetaTrader 5 has a global "Algo Trading" button that can be off after an update, and a bridge cannot trade through it. Fix: put "check the Algo Trading button" on the restart checklist and test with a demo order after every restart.
7. The backtest was too short. Forty trades on one good year is not evidence; it is a story. Fix: insist on a hundred or more trades across different market conditions before you automate anything, and treat a profit factor near one as a coin flip with fees.
8. The settings were tuned until the curve looked perfect. The bot memorised the past and has no idea about the future. Fix: change each setting a little and see if the result collapses; a real edge survives small changes, a curve-fit one does not.
9. Costs were left out of the test. Default strategy testers often assume zero commission and zero slippage. A strategy with 500 trades and a thin edge loses to spread alone. Fix: put realistic commission and slippage into the tester and re-read the profit factor.
10. The stop is in the chart but not in the order. The strategy drew a stop line, the alert did not send it, the bridge placed a naked order. Fix: the stop price travels in the alert message or is computed by the bridge, and the demo test confirms that every order arrives with a stop attached.
11. The bot and the human both trade the account. You see a "better" entry, take it by hand, and now the bot's exit closes your trade or doubles the position. Fix: one account per bot, and a separate account for manual trading if you must.
12. There is no kill switch. When something goes wrong at 2 a.m., the fix has to be one tap on a phone: close everything, disable the bridge. Fix: set it up before the first live trade and test it on demo, so you know it works before the night you need it.
Two habits prevent most of this. First, demo for two to four weeks and compare every bot action with the backtest, so that the surprises happen with pretend money. Second, a fifteen-minute daily routine: check the heartbeat, check open positions against the log, check that yesterday's trades match what the rules said, and check the server has disk space and the terminal is connected. It is dull, which is the point; dull is what a bot is supposed to be.
The full version of this list, with the exact settings for each fix and a printable go-live checklist, is chapter twelve of From Manual Trader to Automated Trader.