Mid-Price Prediction in a Limit Order Book

Abstract

We propose several nonparametric predictors of the mid-price in a limit order book, based on different features constructed from the order book data observed contemporaneously and in the recent past. We evaluate our predictors in the context of an order execution task by constructing order execution strategies that incorporate these predictors. In our evaluations, we use a large dataset of historical order placements, cancellations, and trades over a five-month period in 2013-2014 for liquid stocks traded on NASDAQ. We show that some of the features achieve statistically significant improvements compared to some standard strategies that do not incorporate price forecasting. For the two features that achieve the best performance, the trading cost improvement is on the order of one basis point, which can be economically very significant for asset managers with large portfolio turnovers and for brokers with considerable trading volumes.