Questions abound about the structures of exchanges, the prevalence of off-exchange platforms, and the infrastructure of trading and execution across a wide range of securities and assets. What is more, the cauldron in which these questions boil and bubble gets hotter every time there is some high-visibility wood brought to the fire such as (just to take examples from recent days): a false report on a hacked AP Twitter account that causes the (brief but scary) loss of $136 billion in market cap; Australian authorities jumping on the bandwagon of regulatory suspicion of high-frequency trading and/or dark pools; and a software glitch that shut down the Chicago Board Options Exchange for more than three hours.
Another recent headline that may help keep those trading-tech quarrels raw: it appears that New York State prosecutors will proceed with the trial of Sergey Aleynikov on charges substantially identical to those that were at stake in the federal prosecution that led to a conviction in 2011 that was overturned on appeal in 2012. Those charges arose from Aleynikov’s former employment as a computer programmer at Goldman Sachs, and the mixed open-and-proprietary nature of the coding.
With all these issues running through my head, I spoke recently with Greg Wood, the head of algorithmic execution, listed derivatives and foreign exchange for Deutsche Bank Securities. He has more than twenty years of experience in the broader IT area, and recently has been a regular at FIA and FPL (FIX Protocol Ltd.) events.
Not What to Think: How to Think
Mr. Wood made it clear that he didn’t want to address the specifics of the issues or incidents mentioned above – or others. But he gave me some guidance in how to go about thinking about algorithmic trading and infrastructure matters.
One obvious but important point is that systems can fail. If a market participant relies on transaction at a certain exchange to maintain an over-all properly hedged portfolio, and if that exchange goes black for whatever reason, the market participant can suffer. The market participant should be aware of this.
“It is true of all aspects of trading technology, regardless of what part of the transaction the technology applies to – on the buy-side, in-house brokerage technology, at the exchanges or the off-exchange platforms – that there has to be contingency planning around any proposed change. You have to consider what might go wrong, and how that might be circumvented or rolled back.”
We also discussed tick sizes, a considerable bone of contention of late, in the U.S. and elsewhere.
Wood said, “In forex in particular, some venues used a pip [percentage in point], then a half pip, then a tenth of a pip, then went back to a pip again. It is all a process of trial and error.” In broader terms, he expressed the view that controversies over tick size in a range of markets are part of the process of trial and error in creating liquidity, so that over time each market/asset can find a system that best suits the needs of the participants.
Two Broad Accusations about Algorithms
We discussed two of the general charges that seem to be laid at algorithmic trading a lot: first, that it is a futile arm’s race, benefiting the arm’s merchants (both hardware and software, as well as those in the business of selling proximity) but not doing the broader economy, Main Street, any good; second, that algorithmic trading creates the danger of a disastrous feedback loop, as robots on a bad day decide to play off each other’s downward momentum.
To the first charge, he said: “There are some aspects of recent technological developments that look like an arm's race, for incremental and fleeting advantages of one party over another. But this is part of a process that's been going on for a long time, including Rothschild’s notorious use of carrier pigeons to get an advantage in London trading before anyone else knew the outcome of the battle of Waterloo. And over time the resulting technology innovations and improvements have made the process of price discovery more efficient."
To the second charge, the danger of a feedback loop among algorithms, Wood said: “That concept is a legitimate matter of concern if it is taken to the extreme – however there are protections in place to avoid this. Everybody has a responsibility in addressing potential issues: the providers and users of trading software, the exchanges, the off-exchange platforms. All are working through points around software quality controls and market circuit breakers that help address these concerns.”
Given the wild changes in the means of trading just in the last ten years, to speak of no broader span, we can expect continued uncertainties, until perhaps someday Birnam Wood arrives at Dunsinane Hill to resolve them all for us.