Budish HFT Masterclass

Eric Budish’s latest high-frequency trading explainer argues that modern order books reward tiny speed advantages, not just better prices. (nber.org) Budish is a University of Chicago Booth professor whose research focuses on market design, the rules that determine how buyers and sellers meet. In his finance work, he has argued that continuous limit order books create “mechanical arbitrage rents” that trigger races measured in microseconds. (chicagobooth.edu; chicagobooth.edu) A continuous limit order book is the standard market structure where orders sit in a queue and the first order at a given price gets filled first. Budish’s papers say that when public information changes, that queue can turn into a sprint to cancel stale quotes or hit them before someone else does. (chicagobooth.edu; chicagobooth.edu) His 2021 paper with Matteo Aquilina and Peter O’Neill used stock-exchange message data to measure those races directly. The paper found about one race per minute per FTSE 100 stock, a modal race duration of 5 to 10 microseconds, and roughly 20% of trading volume tied to latency-arbitrage races. (nber.org) The same paper estimated that those races account for roughly one-third of price impact and effective spread, and impose a trading cost of about 0.5 basis points. It put the total sums at stake at about $5 billion a year in global equity markets. (nber.org) Budish’s proposed fix is to stop processing orders one by one in continuous time and batch them into tiny auctions, such as every tenth of a second. His research says that design shifts competition away from raw speed and back toward price. (chicagobooth.edu) That framework now reaches beyond stocks because Polymarket also uses a central limit order book, though in a hybrid setup with offchain matching and onchain settlement on Polygon. Polymarket’s developer docs say prices come from bids and asks on the book, orders are signed offchain, and matched trades settle atomically onchain. (docs.polymarket.com; docs.polymarket.com) Polymarket’s docs also show why the venue appeals to trading engineers hunting micro-edges: the order book is public, prices update through an application programming interface, and official software libraries are available in TypeScript, Python, and Rust. The platform says traders can query best bids, asks, spreads, midpoints, and full book depth directly from its central limit order book endpoints. (docs.polymarket.com; docs.polymarket.com) In that setup, the old high-frequency trading lesson carries over almost unchanged: if one system sees a stale quote a fraction of a second before another trader does, the edge is small on each trade but repeatable. Budish’s data found the average race was worth only about half a tick, with profits adding up through frequency and volume rather than size. (nber.org) Budish’s core claim is narrower than “speed is bad.” His papers say the problem is a market design that lets symmetrically observed public information turn into private profits for the fastest responder. (chicagobooth.edu; nber.org)