Guides
Risk Management
Every layer of protection between an AI signal and your balance — what each one does, when it fires, and what you control.
Defense in depth
Risk management in NeuroTrader is not one feature — it is a stack of independent layers, each able to stop or shrink a trade on its own. A trade executes only if every layer allows it.
| Layer | Protects against | Configured in |
|---|---|---|
| Model agreement gate | One overconfident model | Settings → Minimum Model Agreement |
| Confidence gate | Marginal, low-edge setups | Settings → Confidence Threshold |
| Coherence modulation | Chaotic, untrustworthy markets | Automatic |
| Position sizing | Oversized single-trade exposure | Settings → Position Size / Order Value bounds |
| Max open positions | Over-concentration | Settings → Max Open Positions |
| Daily trade cap | Churning and overtrading | Settings → Max Daily Trades |
| Daily loss cap | Compounding a bad day | Settings → Max Daily Loss |
| Trade cooldown | Re-entering the same move | Settings → Trade Cooldown |
| Averaging-down guard | Unbounded loss-chasing | Automatic + DCA settings |
| Protective exits | Losses running unmanaged | Settings → Stop-Loss / Trailing / Targets |
| Kill switch | Everything else | Safety Console, every page |
Entry protection
The agreement and confidence gates are covered in The Decision Pipeline. The key property: they are vetoes, not weights. A brilliant score on one gate cannot compensate for failing another.
Account-level circuit breakers
Position protection
Once a position is open, a dedicated monitor manages it continuously and independently of new-entry logic:
- Stop-loss — the maximum tolerated loss, always armed.
- Trailing stop — ratchets up as the position gains; gives back only the configured pullback.
- Take-profit / trailing take-profit — fixed target, or momentum-riding exit on pullback from peak.
- Partial exits — bank a configured share at the first target, let the rest run.
- DCA safety orders — strictly bounded averaging-down, only within your configured count and sizing.
Position protection keeps running even if you pause or stop the session — stopping the AI never abandons an open position's protective exits.
The kill switch
The Safety Console (floating shield icon, every page) gives you three escalating actions:
Platform-level safeguards
- Non-custodial design — your funds stay on your exchange; NeuroTrader holds trade-only API access. Withdrawal permission is never required.
- Wave-gated rollout — live access activates in controlled cohorts.
- Global force-to-paper — operations can instantly route all execution to paper in an emergency.
- Full audit trail — every decision, gate result, and halt is recorded and visible to you.
What risk controls cannot do
Honest limits: no risk system eliminates market risk. Gaps and extreme illiquidity can cause exits to fill worse than their trigger levels; daily caps bound realized losses per day, not cumulative drawdown across days; and paper-mode behaviour, however faithful, is still simulation. Read the Risk Disclosure — it is short and written to be understood.