We are currently in an era where fast-moving trading machines and algorithms are outpacing long-term investors. It is no wonder that high-street banks and funds are hiring STEM undergraduates to fill in jobs that used to be reserved for Finance students. Now, in the midst of an ever-evolving competitive environment, clients are more demanding, margins are tightening, and ‘quants’ are part of the solution. Today, they are an integral component of many firms’ competitive strategies. Quoting Scott Patterson, “If Fidelity wanted to buy a million shares of IBM, the Bots [programmed and built by quants] could detect the order and start buying IBM themselves, in the process driving up the price and making IBM more expensive. If Fidelity wanted to sell a million shares of IBM, the Bots would also sell, pushing the price down and causing Fidelity to sell on the cheap (p.5)” . But underlying Patterson’s argument is a more profound issue: how can the rest of the non-quant side of the industry escape this trap?
The answer is dark pools. And in the grand scheme of things, this is partially what they are used for. In my last article, I referred to them as “private exchanges for trading large amounts of securities” and concluded by stating that they are just a component of the overall market structure. This time, we are taking the definition one step further, as well as debunking what we mean by “a component of the overall market structure”. Jamil Nazarali (Head of Citadel Execution Services) describes dark pools as a simple crossing engine . To understand what he means by this, it probably isn’t a bad idea to bring back my diagram of the U. S’s market structure:
Let’s focus our attention on the ‘off exchange’ side of the diagram. Over the last three decades or so, deregulatory frameworks have allowed Alternative Trading Systems (ATS) to be created. As the name suggests, they are simply just another method of competing with traditional exchanges (i.e, AMEX, NYSE, NASDAQ) on the basis of ‘order flow’ –a term used to describe the number of orders that can be routed through an exchange to eventually make prices go up and down . There are two types of ATS systems: ECN’s and Cross Networks/engines. The latter is just another word for dark pool – mostly . There are roughly three kinds of dark pools: independent, broker-dealer and exchange based. I say ‘roughly’ because you may come across other types like consortium-owned and aggregator dark pools. Their names are relatively indicative of what they are – independent dark pools are operated by single firms, broker-dealers are operated by investment banks, and exchange-based pools are offered by some over-the-counter exchange platforms . All these types have one thing in common; they use a proprietary algorithm to match buy and sell orders. With this in mind, we can begin to think about some of the mechanics behind dark pools.
When talking about the history of dark pools in my last article, I mentioned that one of the first dark pool trading venues was called ‘after hours cross’. In retrospect, it was so named because back in the 80’s, dark pools only fulfilled orders after the closing bell at 6:30pm EST. An algorithm would match buyers and sellers using the security’s closing price for the session as means of settlement. Much has changed since then – orders are now fulfilled whenever, but the matching algorithm is still vital to match buyers and sellers using a fair price. But what ‘settlement’ price do algorithms consider? In general, algorithms must use ‘lit venues’ (on-exchange) to define settlement prices . Remember my diagram? It’s the left-hand side. However, the way these settlement prices are determined differs depending on the type of dark pool. For example, broker-dealer pools derive their prices from order flow, so there is an element of price discovery involved. Exchange-owned dark pools gather their prices by complying with the NBBO (National Best Bid and Offer) regulation. NBBO states that prices must be determined using the best bid and ask prices available to customers from multiple exchanges . While the NBBO acts as a safekeeping mechanism to ensure investors receive the best possible price when trading, the major drawback is that the system sometimes fails to provide the most up-to-date information, which means settlement prices could, in theory, be wrong . This oddity is typically referred to as ‘stale pricing’ . Stale pricing comes in two forms: first, processing latency – delays caused by the software and hardware driving the calculations and dissemination of pricing data, and second, transmission latency – the time it takes to transmit data between physical locations. Curiously, stale pricing only seems to be beneficial for one particular set of users: high frequency traders.
Let’s put things into perspective – stale pricing is rare, and has a minimal economic effect. While studies on the market impact of stale pricing are quite rare to come across, a study conducted in 2016 by the Financial Conduct Authority (FCA) concluded that the economic impact of stale pricing costs the market roughly £4.2m per year across all UK dark venues . This is a figure that looks irrelevant in comparison with the average daily book equity value of the London Stock Exchange – about £4.1b or so . However, 96% of the time, high frequency traders are on the benefiting side of the trade. A quick side note: high frequency trading (HFT) is an automated trading system that relies on pre-built algorithms to analyze the markets and spot trends in a fraction of a second. These algorithms are so sophisticated, that they can scan markets in a matter of seconds and execute millions of orders to make money on tiny differences between bid and ask spreads . So, if high frequency traders spot pricing latencies, they are likely to use dark pools as a means to trade large volumes of securities and generate a big chunk of alpha.
HFT is bigger than one might think. In the U.S, the number of stocks that are traded this way represents about 50% of the market. However, HFT’s have recently become scrutinised for creating ‘flash crashes.’ For example, in May 2010, the Dow experienced a 9% drop over a 5-minute period, which took 10 minutes to recover . Unfortunately, many of the flash crashes caused by high frequency trading are vectored through dark pools. Hence, HFT’s can give dark pools somewhat of a bad rep with respect to the channeling of flash crashes through that particular medium (dark pools). To an extent, this is not surprising – HFT dominates daily trading volumes, so there’s a chance that day traders experience sudden peaks and troughs in the pricing of securities, indicating that volatility is not always caused by macro developments. Here we come to the disadvantages of dark pools. In his bestseller Flash Boys, Michael Lewis points out that the opacity of dark pools makes client orders vulnerable to predatory trading practices by high-frequency trading firms . Apart from latencies, HFT’s also use ‘pinging’ as a strategy to put out a few small orders in the market to detect larger hidden ones in dark pools. These strategies can often come across as inefficient or abusive, and the lack of transparency in dark pools surely facilitates these strategies. However, we already know that dark pools are designed to limit the market impact of large orders. In addition, they may also accommodate lower costs, since trades executed through dark pools don’t incur exchange fees. The bottom line is this: dark pools offer some unique advantages to large market players. However, their lack of transparency also generates many topics of conversation in instances where high frequency trading is mentioned. No matter what the case is, one thing is certain: dark pools are here to stay – they are an innovation – a response to the need of trading large blocks of shares with minimal market impact.
-  Scott Patterson, Dark Pools: The Rise of the Machine Traders and the Rigging of the U.S. stock market (p. 5).
-