Tuesday, August 7, 2012

Dark Pools

I picked up Dark Pools by Scott Paterson on Friday evening and finished Sunday afternoon. That ought to say something for the book's readability. I recommend it to anyone that trades, especially equities traders.

The structure of the stock market is far more complicated than I ever expected. Trading at Interactive Brokers, I always noticed that the execution venue would vary between different acronyms like ISLAND, ARCA and BATS. I knew that they were ECNs, but I never really understood what linked them together.


The stock market is not an exchange in the sense of a centralized location where all transactions occur. It is more like a listing entity where public companies go to list their shares. Actual trading occurs on any network plugged into the system.
The "exchange" is really a complex network of networks with varying degrees of favoritism shown to high volume clients.

Electronic Communication Networks ("ECN") originally started in the mid 1980s with the idealistic goal of eliminating the corrupt practices of the NASDAQ floor specialists and market makers. These guys were notorious (and later heavily fined) for colluding to artificially widen spreads on stocks for their own profits. ECNs would cut out the hated middle man while reducing errors, increasing transparency and dramatically decreasing execution time.

As one can imagine, the ECNs took off rather quickly. Not only did they offer much faster execution, but they were also about 60% cheaper to execute a trade.

They corruption that ECNs sought to eliminate inadvertently replaced one problem with another. Networks looking for liquidity offered trading rebates for limit orders that added orders into the system. The setup, which came to be known as maker-taker, created perverse trading incentives for participants. The more trades executed, the more the profits would add up.

Firms sought to become something like Walmart is to wide screen TVs. The more you sell, the more you make. The liquidity providers started fighting aggressively for inside placement of the spread to facilitate ever more trades.

The system spun out of control in two ways. It created a computing arms race where firms focused on purchasing cutting edge technology that shaves microseconds off of calculation time. Firms with the deepest pockets could literally buy an advantage through their computing hardware over the average Joe.

More importantly, the system itself bred its own corruption. The ECNs grew addicted to the liquidity fees. The more trades that fired off, the more money they made. They naturally started catering to their most important clients.

How they did it, though, is what sickens me as a trader. The ECNs started creating order types that were effectively secret. They allowed high speed firms to jump in line over retail chumps using vanilla limit orders. The ECNs offered colocation access at exorbitant fees to give the machines an edge. They allowed the creation of dark liquidity where some players could literally hide orders for execution while others displayed theirs 100% of the time. It makes a complete mockery of the idea of a level playing field.

Artificial intelligence played a big role in how the algorithms operate. What encouraged me, however, was how Patterson chose to wrap up the book. Many of the funds covered near the end start out as relatively small fish working with various types of AI to build predictive trading systems. Their initial results appear encouraging.

One of our biggest projects here is to use various models of fractal markets to build an automated trading system on behalf of the client that I go visit in Ireland so often. Andy and I meet on a near daily basis to discuss our genetic algorithm and how best to encourage the network to behave as we want it to. We are about a month from running our first predictive tests with our in house model. Reading this book encourages me that we are blazing down the right path.