Executioners

Execution Boundary

flowchart LR
    ORD[Order Intent] --> EXE[ExecutionModel]
    EXE --> FILL[Filled Order]

Before execution:

• price is intent
• value_frac expresses capital allocation

After execution:

• fill_price, fill_qty, total_cost, brokerage_cost are finalized

Naive Execution Model (Current Implementation)

The current ExecutionModel implements a deterministic, immediate fill model:

• buy orders consume a fraction of available cash
• sell orders liquidate the full quantity
• fills occur at order.price
• no slippage, no latency, no volume constraints

This is intentional and serves as a baseline.

Buy vs Sell Semantics

Buy Orders

available_cash × value_frac
        ↓
   total_cost (incl. brokerage)
        ↓
   net_cost / fill_price
        ↓
      fill_qty

• Cash constraint is enforced here
• Quantity is derived, not specified

Sell Orders

• Quantity is already known
• Execution determines net receivable after brokerage

Determinism Guarantee

Given the same:

• order
• available cash
• execution model parameters

Execution is fully deterministic.

This is critical for:

• reproducible backtests
• debugging strategy behavior
• regression testing

Separation from Portfolio

The Executioner:

• DOES fill orders
• DOES compute costs
• DOES model market noise

The Portfolio:

• DOES NOT simulate markets
• DOES NOT apply slippage
• DOES NOT guess fill quantities

This separation avoids hidden coupling and keeps accounting clean.

Extensibility: Live Trading & Advanced Models

The ExecutionModel is designed to be subclassed:

• live broker APIs
• limit/market order books
• volume‑weighted execution
• latency and queue priority models

The Backtester and Portfolio remain unchanged.

Swapping execution models does not affect strategy or portfolio code.

Typical Market Noises (not implemented)

Typical execution‑level failures:

• insufficient liquidity
• rejected orders
• partial fills
• delayed execution

These belong only in the Executioner layer and must be surfaced via filled order metadata.

Mental Model

Think of the Executioner as a market adapter:

• Strategy says what it wants
• Executioner decides what actually happens

Summary

• Executioner simulates market reality
• Only source of execution noise
• Deterministic by default
• Fully swappable for live trading

This module is the bridge between research and reality.