Rishi Narang has written another article on cnbc.com defending charges of HFT front-running, this one a direct attack on me.  Titled “Exposing the falsehood of a prominent HFT critic’s arguments“, Narang attempts to refute my rebuttal of his original op-ed in which he claimed that statements in our latency arbitrage paper are “blatantly false”.  His latest column is full of misrepresentations, spurious accusations, and flawed arguments.  Casual readers of Narang’s column will have the impression that I made various admissions, accusations, and false statements, none of which I actually made.  This post is an effort to set the record straight.

First, I was very explicit in my rebuttal that I was not referring to any legal definition of front running (and never have—in fact our original paper does not even mention the term “front run”). I specifically disclaimed any qualifications to address legal matters, and noted repeatedly that legality was not the question at hand.  I cannot imagine how I could have been clearer on this point.  Nevertheless, Narang says:

In his rebuttal, Wellman admits there is no front-running, as legally defined.

I defy anyone to find such an admission in that rebuttal.

Next, Narang quotes me saying “it does not matter much what we call it”, in reference to the phrase “in effect jumping ahead of other orders”. He calls me to task for reckless speech, saying it

shows a gross lack of appreciation for dealing with the fallout of frivolous lawsuits, which are real and which are happening now, thanks to the likes of Michael Lewis and Michael Wellman

Accusing me of encouraging frivolous lawsuits on this basis is rather disingenuous. Of course it matters legally what we call things.  But the statement he quotes was in a context where I had just made absolutely clear that we were not talking about legal definitions.

In my rebuttal I gave a simple example to explain how we had defined “latency arbitrage” in the paper. Narang devotes the rest of his column to arguing that this scenario is very unlikely, and that even if it does happen there is nothing wrong with that. The argument about likelihood (attributed in part to his brother the CEO of Tradeworx) is based on assumptions that the exchanges route orders based on direct feeds rather than SIP, that SIP latency is on the order of one millisecond, and that limit orders also tend to be submitted based on direct feeds rather than SIP. All of these assumptions may sometimes be true, but there is substantial public evidence that they often are not. Moreover, focusing on the likelihood of encountering a crossed market at any particular instant is quite misleading, as it is the aggregate of such events over time and markets that is relevant to the tactic’s profitability.

My point in specifying the latency arbitrage example was simply to explain how it is possible that a latency advantage allows an HFT to take a trade that a preceding order would have taken if they were aware of it.  The phrase “in effect jumping ahead of incoming orders” seems to me an accurate English description of this situation, and Narang does not give any reasons that it is not.  (Recall it was use of this phrase that triggered the original dispute.)

A constructive debate on the facts about specific HFT tactics and their implications for market performance could be quite informative to policy makers and the public. Conducting such a debate, however, requires accurate characterizations of the counterparty position.