Hands-On Machine Learning for Algorithmic Trading

Trading costs

This strategy may assume short sales that require a counter-party, hold less liquid assets that may move the market when traded or underestimate the costs that arise due to broker fees or slippage, which is the difference between the market price at the decision to trade and subsequent execution.

The solution includes a limitation to a highly liquid universe and realistic parameter assumptions for trading and slippage costs (as illustrated in the preceding zipline example). This also safeguards against the inclusion of unstable factor signals with a high decay and, hence, turnover.