drawdown-circuit-breaker

---
name: drawdown-circuit-breaker
description: Use when deciding whether new trade risk is allowed today, based on realized P&L, losing-streak cooldowns, and weekly/monthly drawdown limits.
category: Risk
version: 1.0.0
tools: []
---

# Drawdown Circuit Breaker

Decide whether the trader should take new trade risk today, based on account-level realized P&L and recent terminal trade outcomes. This is a recommendation and recordkeeping tool — it does not replace human judgment and cannot enforce broker-side blocks.

## Inputs needed

Ask the user for (or read from any trade history/notes already in the conversation): account size in dollars, realized P&L today / week-to-date / month-to-date, and the sequence of recent closed trades (win or loss, exit timestamp). With no history available, default to `TRADING_ALLOWED` — don't block a user who has no track record yet.

## Default rules

| Rule | Default | Triggered state | Release |
|---|---|---|---|
| Max daily loss | 2.0% of account | HALTED | Next weekday |
| Losing-streak cooldown | 2 consecutive terminal losing trades | COOLDOWN | 24 hours after the latest loss exit |
| Weekly drawdown halt | 5.0% of account | HALTED | Next Monday |
| Monthly drawdown halt | 8.0% of account | HALTED | First day of next month |

Use the trader's local market timezone (America/New_York for US equities) for day/week/month boundaries. Day, week, and month boundaries should be computed against a stated "as-of" point — if the user gives a date only, treat it as covering the full trading day; if they give a specific timestamp, exclude any trade events after that moment. All four thresholds are overridable if the user specifies different values (e.g. a tighter 1.5% daily loss limit, a 3-loss cooldown trigger, a 48-hour cooldown, or different weekly/monthly percentages).

## Evaluation logic

1. Compute realized P&L today, week-to-date, and month-to-date as a percentage of account size.
2. Count consecutive losing trade closes (a win resets the streak).
3. Check each rule against its threshold; if more than one triggers, report all of them rather than just the first match.
4. Where rules disagree in severity, the more restrictive of COOLDOWN/HALTED governs — a triggered monthly halt overrides a merely-active cooldown, for example.
5. If the trade history supplied is incomplete or ambiguous (missing exit timestamps, unclear win/loss outcome), degrade the result to a lower-confidence read rather than guessing — say plainly which figures are uncertain.

## Recommendation states

| Recommendation | Meaning |
|---|---|
| TRADING_ALLOWED | No rule is active; new trade risk may proceed |
| COOLDOWN | Do not open new positions; manage existing positions and review recent losses |
| HALTED | Stop new entries entirely and focus on review until the halt expires |

Existing position management stays a human decision — the circuit breaker exists to prevent new risk escalation after realized damage, not to force liquidation.

## Output

Report: the recommendation; which rule(s) triggered, with threshold vs. observed value and when the trigger releases; current realized P&L (today/WTD/MTD); consecutive-loss count and the timestamp of the last loss exit; the account size and thresholds used; and a one-line rationale a trader could act on without re-deriving the math (e.g. "2 consecutive losing closes triggered a cooldown; avoid new entries until it expires").

## Key principles

1. **Realized damage only** — use recorded realized P&L, not unrealized P&L or cumulative thesis-level fields, for daily calculations. Unrealized drawdown on open positions is a different, separate concern.
2. **Survival first** — the point is to prevent escalation after losses, not to predict markets or grade trade quality.
3. **Advisory, not automatic execution** — informs the decision; never places, cancels, or blocks broker orders.
4. **Graceful degradation** — no trade history defaults to allowed, so a new trader isn't blocked by an absence of data; incomplete or malformed data should lower confidence in the read, not crash the evaluation or force an unwarranted halt.
5. **Complements, doesn't duplicate, position-level gating** — this evaluates account-level damage; a separate pre-trade checklist (entry plan, stop, size) should still gate any individual order even when the circuit breaker itself reads clear.