Original title: Data-Driven Measures of High-Frequency Trading
Gbenga Ibikunle, Ben Moews, Dmitriy Muravyev, Khaladdin Rzayev · August 2026
Public stock market records do not clearly show high-frequency trading, where fast computers trade in split seconds. Common estimates also fail to separate traders who supply liquidity, meaning they offer to buy or sell, from traders who demand it by taking existing offers. Researchers solved this by training machine learning models on private Nasdaq trading data. The models learned to spot high-frequency trading patterns using only public market clues. The team then mapped these trading styles across all U.S. stocks from 2010 to 2023. They found that supply-side computer trading is linked to better information and lower trading costs, while demand-side trading shows the opposite. During COVID-19, high-frequency supplied liquidity stayed strong.
You are watching a crowded marketplace from above, but the video is too blurry to see which stalls are being set up and which goods are being taken. Without knowing who is offering and who is buying, you cannot tell if the market is healthy or vulnerable. In this paper, the researchers use perfect footage from one market to train a model that recognizes these two roles from blurry public footage of any market. This allows them to track offering and buying activity daily for thousands of stocks over many years.
For decades, researchers could only guess whether high-frequency traders were helping or hurting markets. This work gives them daily data on which kind of HFT is happening in every U.S. stock, making it possible to trace its effects on spreads, information, and resilience during crises like COVID-19.
The approach depends on proprietary data from a single exchange for training, so its accuracy on other platforms not tested is unknown. Validation is limited to one other exchange (Euronext Paris). The daily measures are estimates, not exact trade labels. The reported associations do not prove causation; other factors could drive both HFT activity and market quality.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Data-Driven Measures of High-Frequency Trading — Gbenga Ibikunle, Ben Moews, Dmitriy Muravyev et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.