The Gateway to Algorithmic and Automated Trading

When ghosts go fishing

Published in Automated Trader Magazine Issue 13 Q2 2009

Dark pools have been part of the trading experience, on both sides of the pond, since fragmentation began. But do we have a clear idea of how to deal with them? We do now. Helen Sanders emerges from the shadows with news of the latest bright ideas.

God does not play dice with the universe; He plays an ineffable game of his own devising, which might be compared, from the perspective of any of the other players, to being involved in an obscure and complex version of poker in a pitch dark room, with blank cards, for infinite stakes, with a dealer who won't tell you the rules, and who smiles all the time.

Terry Pratchett, "Good Omens"

Figure 1: Types of Dark Liquidity Pools

Public Crossing Networks. These are "traditional" dark pools formed by agency-only brokerage firms and to which most buy side firms are connected, including POSIT®, LiquidNet and NYFIX Millenium. Most of these cross buy and sell orders (typically at midpoint) but without displaying these buy and sell interests. In addition, some cross trades based on advertisements, such as BLOCKalert, Liquidnet and Pipeline. are some of the most popular advertisement-based pools. An alert usually goes out to the traders (i.e. those with cross-eligible shares on their blotters) but there can be differences between pools.

Internalisation Pools. These pools are established by brokers, initially started by Goldman Sachs (Sigma X) and Credit Suisse (Crossfinder), which therefore include the brokers' proprietary flows as well as external liquidity. Encouraged by cost savings (by crossing orders internally), commission generation (offering dark pool liquidity directly to buy side customers) and alpha generation (interaction with proprietary flow) many brokers have now established these pools, enabling them to market the firm's brokerage services more easily to buy side institutions. Some of these pools are appointing "liquidity partners" to supply or take liquidity from these pools, creating the potential for information asymmetry which can contribute to pools becoming 'toxic'.

Ping Destinations. These are unique pools, such as Citadel and GETCO, operated by hedge funds or electronic market makers which accept only Immediate or Cancel (IOC) orders, and customers' flow only interacts with the operator's own flow. They operate using black box quantitative models to determine whether the pool should accept or reject the IOC order. The major direct customers are sell side firms using dark pool aggregators or smart order routers to "ping" them-hence, their name. These pools are typically priced competitively. However, unlike other pools, these dark pools do not always give mid-point execution) creating a spread. Ping destinations can also discriminate between customers in awarding business, based on criteria such as the nature of customer flow, pricing, speed of incoming order, etc. Such pools have particular applicability to algorithmic traders.

Exchange-Based Pools. Also known as 'hidden pools', pools are either registered by exchanges (e.g. ISE Midpoint Match, Nasdaq Cross, and NYSE Matchpoint) or created as a result of hidden order types supported by ECNs and exchanges (e.g. Nasdaq Cross and NYSE Matchpoint). The reason for establishing these pools is to attract more liquidity to the exchange/ECN. Hidden orders usually interact with regular displayed orders i.e. hidden and regular orders may be crossed. These pools have proved particularly successful due to their pricing and liquidity levels.

Consortium-Based Pools. These pools, such as LEVEL and BIDS, are established by a consortium of brokers most of whom will have their own dark pool; therefore, these pools are effectively 'secondary pools' with highly competitive pricing and therefore significant market share.

(with acknowledgements to Investment Technology Group)

Dark liquidity pools have grown in number and depth in recent years, with over 40 dark pools forming since 2002 and substantial annual growth. While dark pools are better established in the United States, Europe is catching up and will soon reach double figure market share. As Ali Pichvai, Chief Executive Officer at Quod Financial explains, "Dark liquidity pools have become the 'topic of the moment'.

Hitesh Mittal

Hitesh Mittal

Until Q3, 2008, dark pools did not represent a substantial market share. Since Q4, there has been a huge increase in dark pool volumes. Although the percentage of European dark pool market share can still be measured in single digits, compared with around 17% in the United States, there is strong momentum."
But how can dark pools be used to advantage by algorithmic traders? As Hitesh Mittal, Managing Director, Head of Algorithmic Trading, Investment Technology Group, Inc. (ITG) summarises, "With their promise of liquidity and low-impact trading, dark pools have quickly become a mainstay of buy side trading desks."

With most algo traders now at least dipping their toes into dark pools, it is important to understand how they work and the differences between them. In many respects, off-market or "dark" liquidity has similar characteristics to the "lit" market, with similar order types, pricing and prioritization rules. However, as liquidity is not advertised, there is no market depth feed. Dark pools tend to prefer not to advise of trades to any public data feed, or to delay this as long as possible. The aim of doing so is to reduce the market impact of any trade. Different types of dark pools exist with varying characteristics, as fig 1 illustrates.

Approaches to Trading in Dark Pools

Ali Pichvai

Ali Pichvai

With variations between dark pools, algorithmic traders need to be clear whether a dark pool has been formed from brokers' order books and other off-market liquidity, and how liquidity numbers were calculated. For example, some venues count both sides of the trade, or even count liquidity that was posted but not filled. This adds to the complexity of algorithmic trading in dark pools and the need for sophisticated trading tools. For example, as Bruce Bland, Head of Algorithmic Research, Fidessa illustrates, "It is now almost a requirement for algo traders to seek liquidity through dark pools as well as in the lit markets, which demands a different set of capabilities in smart order routing tools. As the dark pool market is developing fast, there are still challenges in developing appropriate algorithms for smart order routing to incorporate them.

Ali Pichvai, Quod Financial, concurs, "Algo traders don't stay in dark pools for more than a few seconds or minutes, depending on how often the dark pool is updated. To sweep these dark pools manually is error prone and far too labour-intensive, leading to a strong demand for automated smart order routing."
Smart order routing is a vital element of algo trading in dark pools, as we will cover more fully in the next edition of Automated Trader. One of the problems for algo traders is that they need to establish order routing
rules not just based on dark pools but in "lit" venues too. Ali Pichvai, Quod Financial continues,
"Usually, an algo trader cannot trade exclusively in dark pools and has to use them in conjunction with other trading venues. Traders need to define a clear process, such as how long to stay in a dark pool. Finally, they need the right technology to implement the strategy and avoid execution risk."

Bruce Bland, Fidessa explains further,"There are various algorithmic approaches to trading with dark pools. The first is to use 'dark only' algorithmic models to sweep and post liquidity directly into dark pools. The second is to use dark enabled VWAP/POV algorithms to additionally sweep pools prior to trading aggressively on lit markets, which improves execution performance."

Problems with Dark Pools

The problem with dark pools is that without visibility of liquidity in the pools, traders have to develop their own strategies for establishing liquidity levels and pricing. Bruce Bland, Fidessa outlines one option,
"The dark models will sweep all pools to which they have access either sequentially, or concurrently by splitting orders across pools. Models will select which dark pools to sweep first based on the probability of sufficient liquidity levels. Using the analogy of a car park for example, you can determine roughly how large a car park is by the number of cars leaving. Although not an exact science, the same mathematics applies to the number of executions from a dark pool indicating which pools to sweep first."

As Bland indicates, however, using post-trade information to determine the liquidity in a pool can be misleading. Hitesh Mittal, Head of Algorithmic Trading, Investment Technology Group, Inc. (ITG) explains, "There are two common misconceptions about dark pools. Firstly, the majority of traders subscribe to the theory that dark pools are truly dark. Second, traders believe that dark pool trading does not impact stock prices."

Bruce Bland, Fidessa agrees, "Although the aim of trading in dark pools is to minimise the market impact, the process of placing an order in a dark pool itself indirectly affects the market price by creating invisible support levels." There are two major issues here: firstly, dark pools leak residual order information; secondly, information leakage, as well as the order process, has an impact on price.

Bruce Bland

Bruce Bland

Information Leakage

While dark pools do not make information until the trade has been completed, every trade creates market signals, eg:

• Post-trade information will reveal what name has been trading, which suggests additional liquidity may be available;
• The trader who has completed the trade will recognize the potential for more liquidity on the opposite side of his trade;
• As most firms advertise completed crosses on Autex and Bloomberg, the location of these crosses is also shown.

While in theory, such leakage should not be harmful, more significant information leakage takes place using forms of "gaming" such as fishing and information sharing. As Bruce Bland, Fidessa discusses,
"'Fishing' is a problem in dark pools with 'gamers' putting in small orders to establish residual order size."

So fishing is a type of echo sounder into a dark pool. Taking advantage of this, 'gamers' are traders who try to manipulate orders in dark pools for their own benefit. Hitesh Mittal, ITG goes on,"Theoretically, getting fills on small orders is not necessarily an indication of residual liquidity. Yet this assumption falls short because dark pools represent institutional liquidity [which]..tend to be large…As a result, if a gamer gets a few fills from a pool, he can generally assume that there is going to be more behind it. Once he locates those orders, he can manipulate the price in his favour."

Another way in which information is leaked is through information asymmetry. In this instance, when a trader sends an order to a dark pool, the information about the order is revealed to the trader or trader's system on the other side of the trade, giving the opportunity to trade against the order.

Price Impact of Information Leakage

While the impact of a trade in a dark pool will be less than that of a "lit" market it is very unlikely to reduce it to zero. In particular, the liquidity that crosses with has to come from somewhere. Some of the orders will come from the public market, as automated broker systems intercept market-bound orders and instead cross them with traders in the dark pool. This disappearance of the opposite side liquidity as it leaves the market will cause an impact. In addition, a trader's order will slow down the market movement in the direction favourable to the trader and speed it up in the opposite, unfavourable direction. In other words, there is a trade-off: reduce the speed of execution by only crossing with dark liquidity or increase it and increase the market impact of a trade.

Trading on the Opportunity

These issues create problems in the pool, often known as toxicity. Gamers can turn these issues to opportunities and manipulate prices. There are various ways of doing this:

i) Gaming with Fishing
There are four elements to this process:

• Selling a few small lots to determine liquidity levels (fishing)
• Rapid purchasing of lots to push up stock price
• Sending a large sell order to the dark pool at a higher price

• Trader pulls back from the dark pool, and prices revert

ii) Manipulating the Midpoint
Alternatively, a trader can fish with a small sell order to find buy orders. Then the seller can send a buy order at a small margin away from the midpoint NBBO stock price followed by a sell order at a small margin away on the other side, therefore manipulating the mid price,. The trader is then likely to withdraw his buy order so the stock price is restored.

iii) Market Making inside Dark Pools
Hitesh Mittal, ITG explains, "In the displayed market, it is difficult to make markets because displayed orders help competing market makers detect their strategies. Dark pools can be a perfect place for market makers to enter or exit a position. When entering a position, they do not need information on other orders in the pool since they can blindly send passive order limits to these dark pools, knowing that the orders will execute when there are market swings." On the other side, to exit a position, a market maker will fish to find information about other orders.

Challenge or Opportunity

The potential for gaming is either a challenge or an opportunity for dark pool participants, depending on how a trader is seeking to use a dark pool. As Hitesh Mittal, ITG concludes,"The effects [of gaming] on dark pool customers are difficult to quantify in terms of profit and loss, but it is generally safe to say that this is not the kind of flow interaction dark pool customers desire. The dark pool is not longer a level playing field and market makers benefit from it more than the traditional dark pool customers."

With liquidity levels in dark pools still growing, there is inevitably a 'saturation' point where the "lit" market can no longer support a dark market beyond a certain size, which depends on it for pricing. As yet, the point at which this saturation occurs has not been tested. Furthermore, as the dark pools grow, they are likely to be regulated more stringently, potentially affecting the reasons why they are attractive to traders.

As Bruce Bland, Fidessa explains, "From a compliance perspective, trade reporting in dark pools needs to be tightened up, as under MiFID, trade reporting rules for dark pools are not as explicit as they could be."
In the meantime, these locations offer substantial opportunities for traders, if you know how to play the game.