Dark pools are a mainstay in the traditional finance markets, but what does that mean in the cryptocurrency context?
From the CMC editorial desk: One of the more confusing, or lesser-known, terms include “dark pools”, for sure. We were fortunate to have LXDX come and give us the low-down on what a dark pool really is, and how it applies in the crypto context. Prepare for new insights in this illuminating (ha!) post.
Defining dark pools
Dark pools have nothing to do with the dark web nor the black market and are a vastly misunderstood part of traditional financial markets.
To break them down a bit, first, forget the word pool. That’s just finance-slang for a market. A dark market is a trading venue that does not display bids or offers (asks). Bids or asks certainly exist, but inside a dark market, they are always hidden. (Any market that displays its bids and asks, by contrast, is termed lit.) Next, the replacement of the word exchange with pool is to differentiate the particular set of users from the larger available liquidity. But the key concept is darkness.
Darkness represents merely the absence of visible bids and offers—it does not imply the absence of trade reports. In many regions, trades must be reported. In the US, trades must generally be reported to the appropriate Trade Reporting Facility (albeit with a delay).
When dark pools are commonly discussed, particularly within the cryptocurrency context, the assumption is that these dark markets would operate as continuous non-displayed limit order books. However, in the traditional markets world, a small minority of dark pools operate in such a manner.
More commonly, dark pools derive their pricing from lit exchanges, and much of the volume trades at the National Best Bid and Offer (NBBO) mid-price. To be explicit, you do not need to specify a price when you seek to buy 10,000 AAPL shares; you simply specify that you’ll buy them at the current NBBO mid-price. Such execution mechanisms are termed tied markets.
In cryptocurrencies, this is, as of today, a bit of a challenge. While there are many lit exchanges from which to draw pricing to form a consensus price index, there is still considerable debate concerning how fair these prices are. This improves day by day, but it may be many months until there are strong NBBO-like prices for coins outside of BTC and ETH.
The necessity of dark pools
Dark pools were initially built to minimize the market impact of displaying institutional-sized orders on lit platforms. Moving trades off-exchange was not new; institutions have always worked to execute trades without tipping off the markets. Much of this volume traded in the call-around markets (literally, by phone). Dark pools promised better liquidity, heightened privacy, and greater liquidity through aggregation of client demand. All without having to pick up the phone.
The popularity of dark pools in the traditional markets has surreptitiously exploded over the last decade. Estimates place volumes growing from roughly 4% of equities volume in 2009 to hovering around 18% of equity volume in US equity markets. More shares trade in dark pools today than on the lit NYSE.
Much of this flight-to-dark trading is a response to the increasing prevalence of aggressive algorithmic strategies. Utilizing multi-factor machine learning models, short-term alpha capture strategies seek to profit by identifying directional buy/sell interests (known as flows) and racing to get in front of the price dislocations such flows will generate. If the intention of the institutional trader is to get the best price on a larger execution, these participants make it significantly more difficult.
Algorithmic traders, while often characterized as parasitic, are simply a response to the mechanisms by which lit exchanges operate—a natural response to the lit ecosystem. By providing the entire market with data on every order being placed, and providing the data at a pace well beyond human reaction (and certainly cognition!) times, the structure of the market, in and of itself, sets up institutional traders to struggle with hiding their intentions.
Front running is often the term used to describe the behavior of aggressive algorithmic traders. The term gets the heaviest use by those marketing dark pool products. As someone who spent their former life designing such strategies, I take moderate offense to the term, despite the fact that I’m here today extolling said virtues of said dark pools. I also hated lemonade until I opened a lemonade stand.
The term “front run” originates from the paper ticket days, where customer orders were taken by hand to the desk or pit (trading floor) and a broker could literally run in front of the clerk carrying the ticket to instead place his own trade. This action, having non-public information about a pending transaction—often a transaction of one’s own client—is clearly despicable. It is, and apologies for the reductionism, theft.
However, statistically predicting that a large order is imminent and trading off that prediction is hardly the same as physically holding the unfair information in hand. Similarly, seeing a large order and racing to grab shares via microwave links to other venues is not front running. Sure, it’s not great for the institutional trader trying to get the best possible execution price, but the algo traders are either reacting to public information or making predictions. The parasites have no privileged information. When they somehow do, it’s a crime.
Problems with trading on lit exchanges with algorithmic traders
Consider that you’re looking to sell a large number of shares of a company. You could take this order to the public exchanges, unloading the shares block by block. You could cannonball (slam the entire order at once) but that could be very dangerous (read: suicidal). More often, sophisticated traders break the order into pieces and execute chunk by chunk, letting the market’s available liquidity refill bit by bit.
Doing so, however, will often still lead to significant slippage (money lost to execution). Each serial execution taps the market’s available liquidity, and every other participant in the market observes your intentions. Moreover, the aforementioned algorithmic participants will detect this naïve flow, react accordingly, and each execution snowballs the next into cascading worse and worse fill prices as both the order and the competition’s trades cannibalize the available liquidity.
One solution is to execute the blocks on lit exchanges more intelligently. One can train models that seek to minimize market impact by hiding the investor’s flow and minimizing the ability for rival traders to detect the directional interest. This is a large industry in of itself. Execution service solutions are a powerful tool; they fragment and obfuscate flows making detection more challenging. As most of these products benchmark their executions to the instantaneous market price, there are strong metrics of their efficacy. That said, they are but one tool, and while amazing for smaller executions, are of only moderate assistance when trying to transact at very large sizes.
Dark trading is another alternative and, for the execution of large blocks, a very strong solution. Trading with a single set of willing buyers, outside of the public market’s view, facilitates a single fair execution price with delayed and dampened price impact.
Simply turning off the lights on the pricing feed is not enough to meaningfully change the market’s dynamics. Traders will probe and fish and find other ways to search for signals. A key facilitator and differentiator in dark pools is that, due to the tradeoffs being made, i.e., favoring bulk vs. instant liquidity, there’s a greater mandate to curate the orders that are allowed into the market. A lit exchange optimizes for getting as many orders onto the trading book as possible; the quality of liquidity in a lit platform is the integral of all near-top-of-book snapshots. This, however, is of little interest to the operator of a dark pool; the pool’s objective function seeks instead to minimize the slippage of large customer orders.
Many behaviors beneficial to a lit exchange’s “liquidity quality” are highly averse to a pool’s. As such, there are often conduct guidelines for participation inside the dark pool. If the only trades you do in a pool are liquidity taking, and perhaps are always arbitrage responses to what’s happening outside on lit markets, you are generally not providing a terribly valuable service to the pool or its customers.
Impact of dark pools
Are dark pools good for the market? Good for retail? Good for institutions? Good for retail because they’re good for institutions? (Who themselves are often mutual funds, ETFs, or other aggregations of smaller investors.)
Yes. Well, probably.
Market fragmentation, the same product trading in many places and on many types of venues, speaks to the core values of blockchain: decentralization, anti-fragility, and less dependence on a few trusted gatekeepers. Different types of platforms are optimal for different types of traders and participants can self-select where they should trade.
Electronic trading is not that old; it dates back only to the early 90s. There’s no particular reason to suspect that the current exchange designs and mechanics are correct. Alternative designs and rules may prove vastly superior. Diversity is good for markets, and the same diversity is good for crypto markets.
As a last aside, the dynamics at the intersection of all these algo traders, retail traders, institutional traders, and traditional liquidity providers are complex. It’s perhaps impossible to attribute cause and effect so clearly in relation to the shifting (meta-game) mechanics that have evolved over the past decade. There are other histories in which we haven’t lived; this analysis is on the one Monte Carlo path we’ve all walked, but there may very well have been many others where events unfolded very differently.
–Josh Greenwald, CEO & Founder of LXDX.co
LXDX is a digital asset exchange platform that achieves the best possible prices by empowering all traders with the powerful technology previously accessible only by elite institutions. Backed by a team with decades of experience in building trading technology and infrastructure, LXDX can execute millions of trades per second across millions of assets, providing performance and scalability noticeably absent from the current digital asset ecosystem. To learn more, visit LXDX.co.
For over a decade Josh has been head of desk and/or a major participant in algo trading
businesses. He led an equities group at DRW before running research for Asia and
heading up KOSPI trading. Josh founded Greenlight Trading and later ran high frequency
for Laurion Capital Shanghai before working on automation and propulsion at SpaceX.
Want to ask Josh a question about Dark Pools? Find him here on Telegram.
 Scott Patterson, ‘Dark Pools’ Face Scrutiny, Wall St. Journal (June 5, 2013)
 Dark Pool as Percent of Total Consolidated Volume. Rosenblatt Securities, Sept. 2015
 Letter from Daniel G. Weaver, Associate Professor of Finance, Baruch College, to Jonathan Katz, Secretary, Sec. and Exch. Comm’n, Nov. 23, 1998
 While active in Korea, we algo-traders were called “super-locusts”
 CME electronic launch in 1992: http://filecache.drivetheweb.com/mr4enh_cme/253/download/Globex+Growth+081008.pdf