Kill switches: the automation rule nobody wants to need
An automated scalping system can lose money at astonishing speed when something goes wrong: a data glitch, a bug, a broker issue, or a change in market conditions. A kill switch is the circuit breaker that stops trading when behavior is abnormal.
Typical triggers include a maximum number of trades per hour, a maximum loss in a rolling window, a maximum spread, unexpected position size, repeated order rejections, or loss of data feed. Any of these should pause the system and send an alert.
Implementation can be in the EA itself, in a separate monitoring script, or on the broker's side through account-level limits. Redundancy is wise, since a bug in the EA may also affect its own safety logic.
Test the kill switch deliberately. Simulate conditions that should trigger it and confirm the response.
Many large losses in automated trading stem from systems running unchecked. A simple stop rule would have limited the damage.
Which kill switch conditions are built into the automated setup, and have they ever fired?
Kill switches: the automation rule nobody wants to need
-
Amanhasnoname
- Posts: 2060
- Joined: Fri Oct 30, 2026 9:45 am
Re: Kill switches: the automation rule nobody wants to need
Two conditions have actually shut the system down in seven months of live trading. Checks run in the expert and in a separate watcher: more than twelve trades in forty minutes, a floating loss of one point eight percent in a rolling hour, spread above three point two pips, three rejected orders in a row, or a feed freeze longer than eight seconds. I keep both because a wild expert might also break its own safety code.
The first fire was a fourteen-second data hole on CHFJPY. The watcher flattened a zero point two lot position; the expert had not yet noticed. The second hit the twelve-trade cap after I left an extra chart on the same magic number, probably saving me from around forty extra looped entries.
The loss cap has never tripped. I test the switches every other Sunday on demo by forcing a wide spread and a dummy rejection streak, because an untested breaker is only a comment in the code. The two live fires were annoying then and useful later.
The first fire was a fourteen-second data hole on CHFJPY. The watcher flattened a zero point two lot position; the expert had not yet noticed. The second hit the twelve-trade cap after I left an extra chart on the same magic number, probably saving me from around forty extra looped entries.
The loss cap has never tripped. I test the switches every other Sunday on demo by forcing a wide spread and a dummy rejection streak, because an untested breaker is only a comment in the code. The two live fires were annoying then and useful later.