“The Myths around Latency Arbitrage”

It was predictable that wading into the public debates around high-frequency trading would attract some attacks on our research.  Nevertheless, it is still novel for me to read a blog claiming that our “research makes a number of basic mistakes”, as was alleged recently by Remco Lenterman, Chairman of the FIA European Principal Traders Association, in a post on the EPTA web site. Mr. Lenterman incorrectly cites a profit estimate we did not make, and attributes to us premises we do not adopt.

The centerpiece of his argument is a horse-race analogy, where he points out that even though people at the track may know the outcome before TV viewers, they cannot profit on this because the race itself happened in the past. As he states (correctly), the track viewers “cannot miraculously make a bet on the winning horse in between the time that they have observed the result and the time that the people at home have observed it”. Of course, this is because betting closes at the actual start of the race. If wagering were instead allowed continuously as the race proceeded, then those watching at even a small delay would be foolish to enter the betting pool after the gun, knowing that informationally advantaged bettors existed.

We are not confused about the relation among past, present, and future. The consolidated data feed at present reflects information about the state of exchanges at some time in the recent past. Knowing the current state of order books at exchanges therefore does indeed allow one to predict the near future of the consolidated data feed. To the extent that any order routing is governed by the consolidated data feed (or indeed any differentially delayed information), the door is opened for latency arbitrage. I am standing by the logic of our study.