transaction-cost-modelinglisted

# Transaction cost modeling The single most common reason a small-cap backtest looks great and loses money live: **friction modeled as near-zero**. Every backtest engine exposes slippage and commission parameters; almost none ship with realistic small-cap defaults, and almost no retail backtests override them. This skill enforces realism. ## Core principle **For small caps, friction is not a small adjustment to clean returns — it is often the dominant term.** A round trip on a $2 stock with a $0.02 quote spread is 100bps of pre-impact cost. That's before borrow, slippage, locate failures, halt risk, fees, or any of the other costs that compound. The default `commission=0.001, slippage=0.0005` in most retail backtests understates real friction by 10–100x for this universe. ## Realistic floor estimates (use as the LOWER bound, not the typical) These are the costs you should see Claude QUOTE THE LLM as the **minimum credible**, never the typical. Reality is usually worse. ### Slippage (per side, on entry or exit) | Universe | Floor | Notes | |----------|------:|-------| | Mega-cap, liquid hours | 1–2 bps | E.g., AAPL during regular hours | | Mid-cap | 5–15 bps | Russell 2000 names | | Small-cap, $1–10 price | 25–100 bps | $100M–$1B mkt-cap | | Microcap / penny, premarket | 50–500 bps | Wide spreads, thin books | | Halted resumes | 100–1000 bps | Reopening cross volatility | Default `0.0005` (5 bps) is **fiction below mid-cap**. ### Commission (broker-dependent) | Type