Trade Signing in Fast Markets

This study assesses the reliability of trade signing algorithms in fast trading environments using the NASDAQ HFT dataset. We compare of several variations of the Leeand Ready algorithm and the tick rule in data that contain true trade signs. Using one second resolution data, the Leeand Ready algorithm outperforms the tick rule and classifies trades with an accuracy comparable to that in prior studies from slower trading environments. Using milliseconddata, the Leeand Ready algorithm has very high trade classification accuracy and shows little performance degradation in subsamples where the market is particularly fast. In robustness testsin TAQ data, we find no accuracy improvement from millisecond timestampswhile the other results hold. We conclude that trade signing is still viable despite the recent evolution towards fast markets, and the use of quote data still increases trade classification accuracy.