If you work on a trading desk, the days all seem to meld into one continuous 6-hour roller-coaster ride. It starts with the anticipation, tension and sudden abruptness of the open. Soon it settles into oscillating blocks of polarity, from calmness bordering on somnambulism to intense, frenetic bursts. Ultimately it finishes with a 30-minute free fall into the close with energy and anxiety intertwining right until the last tick, and then its sewn up with a cathartic, Pavlovian exhalation on the strike of the final bell.
Like amusement park rides, the markets can be exciting and addictive. But ride enough of them and the experiences will eventually dissolve into one.
There was one day, however, that I do remember most specifically back in the late 1990s while working on the program trading desk at ITG. After the bell, I received a call from friend who had just returned from a multi-day training session on a new execution platform called Optimark. The unsolicited call was to tell me that Optimark was too complicated for most traders to adopt and that POSIT, ITGs flagship point-in-time crossing engine, was safe from competition.
My initial thought was, Great! We win! Chalk it up to youthful enthusiasm. Instead of declaring victory, I should have equated Optimarks challenge as the opening salvo in the approaching war over the alternative execution space. As proof, less than a decade later I found myself as a product manager at a rival firm, battling to take a bite out of that same dark-pool market share Optimark failed to capture. Except this time it was alongside more than two dozen other competitors with the same idea.
Dark Pools in 30 Seconds
By now, you all know the story behind the proliferation of dark pools. If not, the 30-second Readers Digest version goes something like this. New regulations in U.S. equities starting in the mid-1990s through the early 2000s − capped by Reg NMS − were designed to create competition between market centers and therefore tighten spreads, lower costs, and create fairer and more transparent markets. These goals were essentially accomplished − insert genial golf clap here − but not without the byproduct of some rather sticky unintended consequences, one of those being the overnight burst in the number of dark pool destinations.
The basic reason we jumped from a few block crossing engines like POSIT and Liquidnet to a multitude of continuous dark pools was simple: The electronic matching of buyers and sellers was essentially regulated into our markets. Prior to the turn of this century, many lucrative careers were made by folks holding phones simultaneously to each ear and negotiating institutional trades.But following the implementation of auto-executable quotes and the formation of an enforceable NBBO, the business of stock crossing had no choice but to move from the black plastic phone receiver to the dark metallic caverns of computer servers.
The Devolution Was Not Televised
Whereas first-generation crossing engines were creative and unique variations of manual block crossing, many post-Reg NMS dark pools were designed to be accommodative (to new rules) rather than innovative. As a result, most of these second-generation dark pools are quite similar in nature and are primarily structured to match a brokers algo/SOR flow against itself or any external orders routed into the pool. It usually fills fewer than 300 shares at a clip.
With many of these dark pools trying to execute these micro-crosses, traders, through complex routing strategies, were forced to essentially chase liquidity across venues. In a general sense, one can argue that in todays markets, interacting with dark pools has devolved into something rather cumbersome and arguably inefficient. Therefore, the main challenge traders face is understanding and navigating across the unintended consequences of those, well, initial unintended consequences.
At this point, there are many tributaries we can slow-paddle down to gain more insight on the matter, with some swampy bogs along the way. Over the course of few Traders Insights columns, I will dissect the corresponding tocks that make dark pools tick. The best place to start is understanding the source of dark pool liquidity, or what we call the inflows.
The Transient Nature of Liquidity
To put some perspective around the liquidity available in dark pools, lets do some simple, back-of-the-envelope math. On average, 7 billion shares trade per day in U.S. equities. Roughly 15 percent, or 1 billion shares, of this volume is attributed to dark pools. If the 20 largest dark pools claim an average 5 billion shares of inflows (a modest estimate for some), that equates to 100 billion shares of dark liquidity available in the market. One billion shares crossed from 100 billion shares available makes for a very modest hit rate of 1 percent for dark venues.
Logically, we know that so many available shares should generate a much higher volume. But that would assume all of these shares are generated from unique orders. Since we know modern dark pools are less like the closed systems of their predecessors, its safe to say that a good portion of the liquidity passing though these pools has been circulated through other venues. This being the case, most dark liquidity is likely procured from two main resources: broker algorithms/SORs and HFT firms.
In the current era of so-called co-opetition, brokers route algo and SOR orders to competing dark pools for many reasons. Sometimes they are encouraged to do so by their larger customers. Other times brokers choose pools that offer some positive or lack of adverse execution performance. Some leverage the economic benefits. Many brokers consider all three in varying weights. Regardless of the motivation, this carousel of order flow between brokers can result in a fair degree of inflow inflation.
The second source of dark pool inflows comes from HFT. As we know, HFT firms generate a disproportionally large percentage of orders in U.S. equity markets today. HFTs are simply interested in profitability and speed. If a dark pool provides enough liquidity and high messaging speeds, most HFT firms will participate. The drop from 100 millisecond latency to sub-1 millisecond across trading venues (lit and dark) a few years back was driven primarily by HFT firms that demanded higher messaging rates. Dark pools were pitched initially as safe havens from speed traders, but many broker pools have been persuaded to allow them in because just a few HFTs could create a major boost to inflows.
Because so much of the same order flow hopscotches between dark pools, it becomes extremely difficult for institutional customers to decipher the best place to seek fills. With this transient liquidity likely outweighing a brokers own institutional flow, the quality and uniqueness of a pool can become obscured. Although recent regulations give customers a better understanding of how each dark pool functions and its general volume profile, the onus still falls on the vigilant consumer, from both a business and best execution perspective, to scrutinize these venues and determine their seaworthiness.
To start, this can be as simple as listening closely to your brokers sales pitch.
Firms that lead straight off with the size of their inflows may indicate that there is nothing unique about their liquidity. Dark pool providers that emphasize spending resources to increase their execution speeds and reduce latencies may care more about appeasing brokers or HFTs than bringing two institutional orders together. Firms that manage their dark pools through their broker-dealer services desk may prioritize their broker relationships over their institutional ones. Where there is smoke, there isnt always fire, but the story is a good indication of firms motivations and incentives.
Beyond what the broker tells you about its inflows, here are some general questions that will help traders probe deeper into their dark pool providers.
Ask a broker for its general sources of liquidity (internal and external) and how much it contributes to the percentage of inflows in the pool. What indications or information is a customer given when your order does execute against a true institutional order? Also, ask for information on your order flow and the controls you have over counter-party filtering. Assume that brokers wont answer all of your questions, especially if they feel other customer information is at risk, but challenge them when that logic seems faulty.
At a very basic level, dark pools are designed to maximize fills relative to the order flow within them. Once you understand the flow, the next step will be to analyze and evaluate matching logic.
We will look more closely at that in Part 2 of Tales From the Dark Side.
– – – – – – – – – – – – –
A 20-year veteran of Wall Street, Craig Viani has been head trader of an electronic trading desk, and built and managed dark pool and smart-order routing technologies. He is currently an independent consultant.
