Risk management
Risk management is the agent’s capital and execution rails. It does not invent entries — the strategy does that. These fields decide whether a signal may fire, how large the order can be, and when the agent must stop.
All three capital rails (Max Exposure, Drawdown Stop, Daily Loss) are required and must be greater than zero. There is no “0 = off” mode.
Defaults below match a typical create loadout: Max Exposure $100, 1× leverage, sizing 1% / 2% / 3%, Min Allocation $10.
Quick map
| Field | One line |
|---|---|
| Min Allocation | Smallest order (USD notional) you will allow |
| Max Daily Trades | Hard cap on executions per UTC day |
| Balance Threshold | Wallet equity floor — below it, no entry evaluation |
| Max Leverage | Ceiling on leverage this agent may use |
| Max Slippage | Cap on market-entry slippage (bps) |
| Max Exposure | Reusable ceiling on capital at risk at once |
| Drawdown Stop | Cumulative realized-loss halt |
| Daily Loss | Per-UTC-day realized-loss halt |
| Small / Medium / Large | % of budget headroom put at risk per trade |
Budget headroom = Max Exposure minus capital already at risk on open positions. When the agent is flat, headroom equals Max Exposure.
Min Allocation (USD)
Minimum order size in USD notional. If the sized order would be smaller than this, the trade is aborted (not placed as a tiny fee-burner). Platform floor is the exchange minimum (typically $10 on Hyperliquid); you can set higher.
Example. Min Allocation $50. A signal that would only size a $30 order is rejected.
Max Daily Trades
Hard cap on how many trades this agent may execute in a UTC calendar day. When the count is hit, further entries are rejected until the next UTC day.
Example. Cap 3. After three fills today, a fourth signal is skipped even if the strategy likes it.
Balance Threshold (USD)
Minimum wallet equity before entry evaluation runs. Below this, the agent does not look for entries. If equity cannot be read, evaluation is blocked (fail-closed) — it does not keep trading on an unknown balance.
Example. Threshold $200, wallet equity $180 → no entry evaluation until equity recovers.
Max Leverage
Upper bound on leverage for this agent’s trades. Effective leverage is the minimum of this setting, the asset’s exchange max, and what the model picks. Default is often 1×.
Example. Max Leverage 5×. Even if the model or coin allows 10×, this agent never exceeds 5×.
Leverage caps how large a notional the headroom can post as margin. It does not multiply the risk % into a bigger dollar loss.
Max Slippage (bps)
Cap on accepted market entry slippage, in basis points (100 bps = 1%). The system uses the tighter of this agent ceiling and a bracket-derived viability cap. Raising the agent value above the platform entry ceiling does not unlock looser fills — geometry binds first. Lower values mean stricter entries (more chance of a miss in fast markets).
Example. 50 bps → every market entry from this agent uses ≤ 50 bps slip. 300 bps (common default) mostly defers to the bracket cap.
Max Exposure (USD)
Reusable ceiling on capital at risk at once — not a daily wallet and not a depleting bankroll. As positions close, headroom returns.
Example. Max Exposure $100. Agent opens a trade that puts $40 at risk → remaining headroom $60 for new risk. Close that trade → headroom is $100 again.
Platform maximum is account-bounded (often up to $250). Drawdown Stop and Daily Loss cannot exceed Max Exposure.
Drawdown Stop (USD)
Lifetime cumulative realized-loss stop. When losses reach this number, the agent auto-halts. It stays halted until you raise the stop or reset the drawdown baseline (then resume).
Example. Drawdown Stop $100. Cumulative realized P&L hits −$100 → agent halts. Open winners do not clear this; only raising the stop or resetting the baseline does.
Daily Loss (USD)
Per-UTC-day realized-loss stop. A breach auto-halts for the rest of that UTC day; the halt clears at the next UTC midnight. Raising the limit also works. The drawdown-baseline reset does not clear a daily-loss halt.
Example. Daily Loss $50. Today’s realized losses reach $50 → halt. Tomorrow UTC, the daily counter resets.
Position sizing (Small / Medium / Large)
These are not “% of the order” or “% of wallet equity.” Each is the share of budget headroom put at risk when the model picks that size preset.
Formula (hard stop = the protective stop resting on the exchange):
notional ≈ (presetPct / 100) × budgetHeadroom ÷ hardStopDistancePct
A wider hard stop → smaller notional for the same dollars at risk. Must stay Small < Medium < Large (typical risk budgets 0.5–3%).
Worked examples (flat agent, Max Exposure $100, leverage 1×)
| Preset | % at risk | Hard stop from entry | Approx. notional | Dollars at risk |
|---|---|---|---|---|
Small 1% | $1 | 2% | $50 | $1 |
Small 1% | $1 | 5% | $20 | $1 |
Medium 2% | $2 | 4% | $50 | $2 |
Large 3% | $3 | 4% | $75 | $3 |
Same $1 at risk on Small whether the stop is tight or wide — only the position size changes.
Headroom shrinks while in a trade. If Max Exposure is $100 and $40 is already at risk, Small 1% risks $0.60 of the remaining $60 headroom — not $1 of the full ceiling.
Min Allocation still applies. If the formula would size below your Min Allocation (or the exchange floor), the trade is aborted.
How the rails work together
- Balance Threshold — may we evaluate at all?
- Max Daily Trades / Daily Loss / Drawdown Stop — may we still take risk today / this book?
- Max Exposure — how much headroom is left?
- Sizing + Max Leverage + Min Allocation — what notional is legal?
- Max Slippage — will the entry fill under the cap?
Create does not start trading. After create, deploy on Radar, Arena, or the trade terminal.