Optimal Trading Algorithm Selection and Utilization: Traders’ Consensus versus Reality

Abstract

We study the usage patterns of traders employing execution algorithms with the goal of statistically minimizing implementation shortfall. Based on widely accepted concepts, algorithms are categorized into scheduled, participation, dark, and dynamic types. We present the consensus patterns that emerge in the trading behavior and offer insight into how to improve performance. Our goal is to help improve implicit trade cost by studying the right context in which traders use algorithms. We use the mean, standard deviation, skewness, and kurtosis of the distribution of trade cost against arrival price to quantify expected cost and risk. We quantitatively characterize the relationship between trade cost and order size, participation rate, average daily volume, algorithm duration, limit price, and algorithm type.

We find that for a given market condition and order requirement, selecting the right algorithm and tuning its parameters could significantly affect execution performance. Implementation shortfall algorithms exhibit superior performance for small orders, and dark algorithms perform better even in high-participation-rate scenarios. Scheduled algorithms are most sensitive to participation rate and order size, and dark algorithms are least sensitive. The use of limit price and lower participation rate in algorithms greatly reduces the unattractive positive skewness of trade cost distribution and thus improves performance. We also offer insight into setting optimal participation rate and limit prices.