Its not often that you feel like you are reading the right book at the right time, but the recent events at Knight Capital1 gave me that feeling with Dark Pools. Early on there is a prediction that someone will lose $1bn on algorithmic trading in 2012 and I think the author of that prediction will be claiming a moral victory, even if his amount hasn’t been hit yet.
Although the subtitle talks about the ‘looming threat to Wall Street’, the majority of the book doesn’t really talk about that. Instead we get a history of electronic trading. Curiously most of the action takes place well away from Wall Street, largely because the vested interests of exchange trading floors are the enemy in this story, trying to stop technology eating their cake. Of course, we now know they failed. The reader is left wondering if perhaps we all pay some of that price, as some sort of orderly transition may have avoided some of today’s problems – a counterfactual that is perhaps more for the debating floor than the history books.
The story itself feels rather narrow, being primarily about Josh Levine and his creation, the Island trading platform. Although it also discusses other companies, particularly Archipelago and Tradebot, an unaware reader might be surprised when Renaissance Technologies suddenly appears as a company that has been doing AI/algorithmic computing for years. Given its reputation as a secretive company the lack of information is no surprise, but I was wondering what else had been left out of the narrative. Having said that, the story is highly readable and reasonably entertaining. Its really a classic story of a hero taking on the big guys and winning, though you do get the feeling that some of those who tagged along for the ride did do better financially than Levine himself.
The really interesting question, of course, is where does that leave us now. Despite the subtitle on the front, Patterson covers this relatively briefly. His story of the demise of Trading Machines hints at a market designed for insiders, and perhaps can only really be appreciated by someone who is also a bit of an insider themself. The coverage of the flash crash of 2010 is clear and concise. It was stopped by a circuit breaker and these have now been introduced as the official protection against a repeat. Though its difficult to have confidence that this is a robust solution, the lack of a repeat since is somewhat reassuring and they do seem to have helped prevent the Knight Capital fiasco affecting the whole market.
When even those who work inside the high-frequency trading (HFT) world are starting to characterise the markets as being in a mess then its clear we have a problem. Spoofing, putting up orders and cancelling them without them being filled in order to draw out other hidden orders, is technically illegal but by some assessments accounts for 90% of orders. Special order types designed especially for HFT allow them to jump the order queue, at the expense of other traders in the market. And with the bots and algos always on the lookout for the whales (large orders which are hidden,) many investors are looking to dark pools to do their trades, reducing market transparency. While spreads may look like they are down to pennies, in practice for more than a few hundred shares they are effectively wider than they were in the old days of the floor. And when the going gets tough, the algo traders disappear and liquidity along with it. The sad truth is that unless you employ equal or better technology then for anyone who trades in any size, and this includes many private investors, you pay the high-frequency guys to trade. And the public is starting to realise this, leading to a loss of confidence in the markets.
So what’s to be done about it? I think there a couple of simple steps, though how much effect they’d really have is questionable. Clearly there needs to be a principle of equal access for all. That means no queue jumping. It may mean an end to exchanges auctioning off bays in their server farms. Those algo traders who are committed to supplying liquidity should have larger constraints on them – minimum size & spread for starters.
But I suspect all that does is scratch at the surface of the problem. The AI genie is well and truly out of the bag and, as Patterson’s last chapter shows, is only going to become more widespread. While getting ahead of the curve might be impossible, regulators and legislators need to at least start catching up and put rules in place that create a level playing field for everyone. Although the investment banks and traders might scream, perhaps then the energy and investment being put into cutting trading times by milliseconds might then be put to more productive uses.
1 A problem with Knight Capital’s algorithmic trading systems led it to lose $440m in half an hour of trading on 1st August 2012. For see the news articles here and here (I know the latter is a bit sensationalist but does give a good sense of the issues too.)