02_risk_position

Prompt file imported from acid0ikario/trade-bot (.codex/prompts/02_risk_position.md). Copyright stays with the author.

You are acting as a senior Python trading infra engineer. Implement the risk management and position sizing modules and their tests for this repo.

Context

• Project: trade-bot (Python 3.11, Poetry). The exchange wrapper is already implemented.
• The existing files src/bot/risk.py and src/bot/position.py are stubs created in the bootstrap phase.
• Tests for these modules do not exist yet.

Goals

Provide robust functions to compute stop-loss levels, enforce daily loss limits, and size positions according to risk.

Tasks

1. risk.py (src/bot/risk.py):

   • Keep the existing compute_stop(entry, atr, k) helper.
   • Add stop_by_swing_low(prices: list[float], lookback: int = 20) -> float:
     • Given a list of historical low prices (recent lows), return the minimum of the last lookback values. If the list is shorter than lookback, use the whole list.
     • Ensure the returned stop is positive; if no data is provided, raise a ValueError.
   • Add max_daily_loss_guard(pnl_list: list[float], base_equity: float, max_loss_pct: float) -> bool:
     • Compute cumulative PnL for the trading day from pnl_list (each entry could be positive or negative).
     • Return True if cumulative loss is less than or equal to -(base_equity * max_loss_pct); otherwise return False to signal that trading should halt.
   • Add kill_switch(pnl_list: list[float], base_equity: float, max_loss_pct: float) -> bool:
     • Wrapper that simply returns not max_daily_loss_guard(...). Helpful for the runner to decide whether to disable trading.
     • Include a docstring explaining its purpose.

2. position.py (src/bot/position.py):

   • Extend position_size(entry: float, stop: float, equity: float, risk_pct: float, step: float = 0.0) -> float:
     • Compute the dollar risk per trade: risk_amount = equity * risk_pct.
     • Compute per-unit risk as abs(entry - stop); if per-unit risk is zero or negative, raise a ValueError.
     • The raw quantity is risk_amount / per_unit_risk.
     • If step > 0, floor the quantity to the nearest multiple of step (similar to exchange lot size rounding). Otherwise, return the raw quantity.
     • Ensure the returned quantity is strictly positive; raise ValueError if it becomes zero after flooring.
   • Update docstrings to describe the new parameter.

3. Tests:

   • Create tests/test_risk.py with pytest:
     • Test that compute_stop returns entry - atr * k and never negative.
     • Test stop_by_swing_low returns the minimum low in the last lookback prices, and that it raises if given an empty list.
     • Test max_daily_loss_guard:
       • Returns True when cumulative losses are within limit.
       • Returns False when cumulative losses exceed -base_equity * max_loss_pct.
     • Test kill_switch returns the logical inverse of max_daily_loss_guard.
   • Create tests/test_position.py with pytest:
     • Test that position_size returns the correct quantity for simple cases (e.g. entry=100, stop=95, equity=2000, risk_pct=0.01, step=0.1 → qty=4.0).
     • Test that flooring to step works: quantity is rounded down to nearest step.
     • Test that passing zero or negative per-unit risk raises a ValueError.
     • Test that a tiny step or large risk_pct does not produce zero or negative quantity.
   • Ensure these tests run quickly and do not depend on ccxt or network.

4. Adjustments:

   • If you need to import additional modules (e.g. math for flooring), add them at the top of the relevant files.
   • Make sure to run poetry run pytest -q locally; all tests should pass.
   • Update the prompts folder only if you change the spec (no changes required here).

Acceptance criteria

• src/bot/risk.py and src/bot/position.py contain fully implemented functions with docstrings.
• tests/test_risk.py and tests/test_position.py exist and cover edge cases.
• poetry run pytest -q passes on all tests.
• Existing functionality in other modules remains unaffected.

Now implement these changes.