Transaction Cost Feasibility

Before you wire costs into a backtest engine, ask a simpler question: does the signal have enough gross edge to pay for trading at all? If not, the strategy should die early.

The Problem

The full cost of a trade has three layers: explicit costs (commissions, fees), implicit costs (half the bid-ask spread paid on every execution), and market impact (your own order moving the price against you). Most bad strategies fail this economic screen before any engine-specific implementation details matter.

A strategy with 50 bps gross alpha and 30 bps round-trip costs has a safety margin of only 1.7x - too thin to survive estimation error.

The Pattern

WRONG

import numpy as np