Man vs Machine: can the City be run by one man and his dog?

I’ve used a joke for a long time about the City being run by just a man and a dog.  The dog is there to stop the man from touching the computers. The man is there to feed the dog.

Image: Serpent's hand

It’s certainly true that computers have taken over much of City trade these days, and this is reinforced by comments from Leda Braga in the Financial Times  today.

According to Ms. Braga, who runs the $8.9bn BlueTrend hedge fund, traditional investment approaches might soon struggle to keep ahead of so-called “systematic” computer models, as human fund managers are undercut by cheaper and more efficient technology.  She claims that humans cannot possibly keep up with the volume of knowledge the systems can deal with, although they may be better at interpretation.

Long term, that’s questionable however, as artificial intelligence becomes ready for prime time.  According to Dave Coplin, a senior Microsoft executive, artificial intelligence is about to outpace human ability.

Mr. Coplin points out that Moravec’s Paradox, a hypothesis put forward in the 1980s that “states that what we think is easy, robots find really hard, and what we think it really hard, robots find easy. Complex maths equations are hard for humans but take nanoseconds for a computer, but moving around and picking things up is easy for us, while being almost impossible for a robot.  It would be hard to train a robot to be a nurse, or even a chef, but the City could be run by algorithms. People who use their hands will have jobs for life.”

Yep, that reinforces the man and dog hypothesis and can be further substantiated when you look at the fact that High Frequency Trading (HFT) accounted for over 60% of all U.S. equity trading in 2010.

However, according to the SEC report Equity Market Structure Literature Review Part II: High Frequency Trading, it’s only certain parts of the market that this applies to. For example, analysing NASDAQ datasets finds that over half of all HFT activity is aggressive, liquidity taking orders that trade immediately against passive resting orders. Moreover, HFT activity varies greatly across different types of stocks, with the report citing that HFT is far less active in small-capitalisation stocks than in large-capitalisation stocks.  That’s because small-caps are less liquid and provide less systemic information for the computers to crunch.

There’s also another issue, in that HFT has been tarnished by accusations of flash trading and dark pool gaming, which has reduced support for such automation.

A flash trade is a trade purely sent to order equities that others are buying in order to raise the price before their orders are fulfilled.  The flash trade is cancelled before settlement, and was never intended to be a buy order.  The message results in the price of the equity rising before it is filled for others, and the suspicion is that market makers send millions of these messages in a millisecond, matched by millions of cancellations in a millisecond.  The result is that the true buyers pay $0.01 or more per equity per transaction, purely as a result of this gaming of the price. 

Combine this with dark pool trading, which represents 40% of all trading in the markets today.  Dark pools allow price discovery to be hidden, in order to allow large order filling for institutional clients without giving away their trading strategies.  However, this market has also been tarnished with the allegation of fraud against Barclays  (an accusation they reject).

Maybe that is why equities trading has tanked generally since 2007.  According to Institutional Investor, equity trading is down 30% between August 2007 and August 2013 with revenues almost half – down 45% - in that period.

But the result of this crash in trading purely means that the Wall Street banks have taken out more human traders to replace them with more machines. But the machines are also losing money.  According to the research company Tabb Group, HFT profits have bombed since 2009, from $7.2 billion to just $1.1 billion in 2013.  This is because, as Bloomberg reports,  “as much as two-thirds of all stock trades in the U.S. from 2008 to 2011 were executed by high-frequency firms; today it’s about half. In 2009, high-frequency traders moved about 3.25 billion shares a day. In 2012, it was 1.6 billion a day. Speed traders aren’t just trading fewer shares, they’re making less money on each trade. Average profits have fallen from about a tenth of a penny per share to a twentieth of a penny.”

Maybe that’s because it’s just a race based upon power that can be copied.  Even though HFT has reduced the arbitrage opportunities from 97 milliseconds in 2005 to 7 milliseconds in 2011, the profitability of the trades hasn't changed according to an academic analysis by Professors Budish, Cramton and Shim.

Where that leaves us right now is that computers can deal with almost every part of the financial ecosystem, except those that cannot be systemised such as illiquid assets or large block trading structures.  That is because computers don’t have opinions and knowledge in these areas that their human counterparts bring to the table, and that’s where markets believe they can make a difference.  But, one day …

 Image: 2001 a Space Odyssey