To Avoid Latency Arbitrage, Trade Multilaterally

Posted on: June 19, 2019

Don Ross

On April 10, Matt Clarke published a superb overview and analysis of anti-latency arbitrage (ALA) mechanisms here in Tabb FORUM, making a strong case for incorporating them (“Speed Bumps: Anti-Latency Arbitrage Mechanisms Make Markets Better”).  I share his view that the high (and rising) cost of shaving nanoseconds in the race to take liquidity before others is a form of rent-seeking and a drag on financial-sector productivity.  To me, however, the important question is not whether ALA can help mitigate these costs (it can), but whether we can improve market structure such that liquidity arbitrage is itself a meaningless activity (we can).

The bedrock mechanism for exchanging securities has for decades now been continuous trading in an electronic order book.  Matt’s piece, understandably enough, takes it as given that it will continue to be.  To me, however, this mechanism is anachronistic, and we should recognize that its limitations are the root of the liquidity-arbitrage problem.  Who could believe, after all, that in a market dominated by institutions needing to buy and sell tens or hundreds of thousands of shares that a structure yielding 150-share transactions was optimal?

If we accept that it’s not, what would a better system look like?  Well, to transact an average of tens of thousands of shares per trade without moving the spread participants need to transact multilaterally and simultaneously, rather than bilaterally and sequentially—as they do now in continuous markets.

What sort of system can achieve this?  Well, an auction, naturally.  Auctions are already an important feature of the markets, with the NYSE close now representing over a quarter of daily volume in its listed stocks.  But such auctions have a clear limitation—having to wait around for the event, not knowing how the markets will move during the hiatus.

That’s where better technology helps.  Better technology means doing auctions on-demand, whenever traders want them.  Immediately.

But shouldn’t a trader initiating an auction mean information leakage?  Not if you reveal nothing beyond the security you want to transact in.

CODA Markets runs on-demand auctions with symbol-only alerts.  Take a look at this graphic below.  On this particular trade, in small-cap Core-Mark Holding Co (“CORE”), the trader initiated an auction on April 18 at 1:14 PM.  30 seconds later, he had himself 24,600 shares at the NBBO midpoint.  That’s over 240 times the average trade size for the stock, with virtually no market impact.

Note that there was no need for ALA, as no one traded before anyone else.  With the intelligent design of the auction, all participants in the trade transact simultaneously.  On-demand auctions do not need “speed bumps,” or indeed any restrictions on HFT.  They bring supply and demand together at a single point in time, without selling anyone a speed advantage or inserting mechanisms to take them away (which are themselves game-able).  Speed is simply irrelevant.

When electronic order books started taking over the markets in the ‘80s and ‘90s, there was not enough computer power to drive on-demand auctions.  But that limitation is now a thing of the past.  Auctions like this can be run all-day long, for thousands of securities.

As long as the continuous-trading architecture persists, the demand for speed and protection from speed  will persist.  The arguments about LA and ALA will therefore go on as well.  The winner, in my view, will be those who sit it out and trade multilaterally.

Don Ross is CEO of PDQ Enterprises, the parent company to CODA Markets.