Freqtrade kill-switch searches overlap with Freqtrade stop loss strategy bot and kill-switch trading bot because operators want a pre-written rule that stops trading when reality diverges from research.

A kill-switch is not a single setting. It is a policy: under condition X, you stop trading and review.

When to stop

Drawdown beyond a pre-agreed band

Size the band from research and account size, not feelings (Position sizing).

Live slippage and fees worse than modeled

If freqtrade backtest vs live performance diverges beyond cost stress, pause (Why live fails).

Operational failures

API errors, repeated partial fills, or data feed gaps.

Strategy drift

If your live metrics diverge from walk-forward expectations without a clear causal story, pause and reassess (WFE).

How to implement in practice

Use exchange-level stops and bot-level stops together
Log stop reasons so you do not rationalize after the fact
Prefer reduce size before full stop when uncertainty is operational, not statistical

How Kiploks complements bot controls

Kiploks does not replace your exchange stop button, but it helps you justify thresholds using structured evidence before you go live (Verdicts).

Practical Freqtrade operator habits

keep a dry-run mode checklist before enabling real orders
store config.json next to each live session log
alert on order reject rate and cancel/replace loops, not only PnL

Most catastrophic losses start as small operational anomalies that were visible early.

Escalation ladder (reduce, pause, flatten)

Default policy:

reduce size when drift is uncertain but suspicious
pause new entries when drift is confirmed
flatten when risk of further execution damage dominates upside

Write thresholds for each step in advance.

Pair exchange protections with strategy logic

Exchange stops help, but they do not replace bot-side pauses when data is bad.

Treat exchange and bot controls as two layers, not duplicates.