For years, buyside traders have complained – quite fairly – that dark pool trading rules and activity are far too opaque. In response to this criticism and to the numerous regulatory actions taken against dark pool operators, regulators have attempted to shed a light on dark pool operations. Most recently, the SEC made it mandatory for each dark pool operator to file a Form ATS-N, which provides specific, detailed information on the inner workings of the pool. The completed forms are now available on the SEC website.[1]
In this blog post, we use the information gleaned from the completed Form ATS-N’s to provide insights into how dark pools differ and discuss how these differences can be used to enhance execution performance. For the most part, the focus of this post is on “firm order” interactions in the pools, but the general themes noted in the post apply to conditional orders as well.
Differences across dark pools
Priority rules
Most dark pools adhere to price/time priority. But this is by no means uniformly followed across dark pools. Size is occasionally used as a secondary priority rule in place of time, as is the case with MS Pool and Virtu’s Posit. Posit, for example, allocates executions on a pro rata basis, allowing larger orders to receive greater allocations than smaller orders, even if the smaller orders arrived earlier.
Other dark pools use “broker” as a secondary factor ahead of “time”. Goldman Sachs Sigma X2, for example, utilizes price/broker/time priority, which allows a routing broker’s own client orders to interact with one another before interacting with orders from other brokers. JP Morgan’s JPM-X pool uses a price-tier-time priority mechanism, where the tiers favor JPM’s own clients. An incoming buy order, for example, may interact with “Tier 1” clients, such as their own institutional clients, before interacting with orders sent by other brokers. This particular priority system gives JPM institutional clients a distinct advantage over other participants in JPM-X since JPM clients effectively move to the front of the queue, ahead of other users (at the same price).
Tiering
Dark pool operators often tier their clients based on various metrics. Tiering can be done by some objective characteristics of the client, such as the client type (e.g., institution, retail, etc.). Tiering can also be done on the basis of past dark pool performance. For example, a common approach is to measure the “adverse selection” of each client’s trades to measure the “toxicity” of their order flow. Users can then choose which tiers they wish to interact with, where clients in worse tiers are typically restricted from trading with higher-tiered clients, unless those higher tiered clients “opt in”.
Note that while tiering has become fairly widespread over time, even this approach is not uniform across pools. Not only does the tiering mechanism vary dramatically across dark pools, some dark pools, such as Posit, do not tier clients at all. Rather, Posit attempts to limit adverse selection by allowing clients to “block” certain counterparties on request, e.g., because of a bad trading experience with that (anonymous) counterparty. Posit also utilizes a tool called “Liquidity Guard”, that aims to prevent bad fills from occurring.
The ability for a broker’s own clients to gain priority and to more finely calibrate potential counterparties within the broker’s dark pool provide a justification for biasing routing toward the broker’s own dark pool. For example, when using a broker’s VWAP algorithm, the algorithm may rationally choose to preference the broker’s own dark pool because the algorithm’s child orders receive better priority and/or have greater control over their counterparties relative to other dark pools. But brokers also have strong economic incentives to bias their own pools, specifically because they avoid the commissions paid to other venues. Therefore, whether biasing a broker’s own dark pool actually enhances execution quality is, in the end, an empirical question, one which can be assessed through post-trade analysis.
Price Improvement
Most pools, like exchanges, execute trades based on the price of the resting order, implicitly giving any price improvement to an incoming order. For example, an incoming market order that executes against a midpoint peg order will generally receive the midpoint price, even though the market order was willing to pay a price in excess of the midpoint. While this pricing approach is the most common in financial markets generally, some venues choose to give price improvement to the resting order. Credit Suisse Crossfinder is an example of a venue that allocates price improvement to the resting order. The incoming market buy order mentioned above would execute at the offer price, even though it executed against a midpoint peg order, thereby providing price improvement to the resting order.[2] The possibility of price improvement provides an incentive to traders to rest liquidity on Crossfinder (though this also disincentivizes liquidity takers from sending marketable orders, preferring instead to send midpoint pegged orders to CS Crossfinder or to route their marketable orders to other venues with more favorable pricing rules).
Pegging instructions
Dark pools provide various pegging options, e.g., traders can peg the near quote (“primary peg”), the midpoint (“midpoint peg”), and the far quote (“market peg”). But not all venues provide this richness of pricing for all order types. Some venues may restrict trading to the midpoint, as is the case for conditional orders routed into some dark pools. Clients wishing to capture spread or those willing to sacrifice spread to complete the trade would need to look elsewhere for this functionality.
Immediate-or-Cancel (IOC) orders
Most venues allow its users to send immediate-or-cancel (IOC) orders. But some venues, like Morgan Stanley, restrict IOC usage in their dark pool. Specifically, only Morgan Stanley’s own smart router is allowed to route IOC orders in to MS Pool, which gives Morgan Stanley the ability to tightly control IOC usage to only those tactics it views as acceptable. Such restrictions prevent IOC “pinging” aimed at gleaning information about resting orders as well as other extremely short-term “predatory” trading strategies (e.g., latency arbitrage). Controls to prevent information leakage are particularly valuable to dark pools like MS Pool that use size priority, since these types of behavior undermine the very incentives size priority provides users to post large orders.
Multiple pools/benchmarks
While most brokers operate only one dark pool, some brokers operate multiple pools. Morgan Stanley, for example, not only has its “main” pool (MS Pool), but also has a retail pool (MS RPOOL) and a trajectory cross pool (MS Trajectory Cross). The retail pool allows retail clients to interact with one another as well as with certain MS institutional clients. The trajectory cross is a matching system that pairs off algorithmic orders and executes them at the VWAP over the matching period. In the case of Morgan Stanley, the “other” pools – MS RPOOL and MS Trajectory Cross – are available only to MS clients.
Special Features
Perhaps the most famous example of a dark pool “special feature” is IEX’s “magic shoebox”. This mechanism slows orders by 350 microseconds to protect orders from latency arbitrage. IEX alos utilizes a proprietary “crumbling quote indicator” that allows its Discretionary and Primary peg order types to be repriced at less aggressive levels if it appears that prices are highly likely to move. For example, if a downward price move appears imminent based on the crumbling quote indicator, IEX lowers the peg price on buy orders to prevent them from being “picked off” by high-speed traders.
Another “special feature” offered by some dark pools is periodic instead of continuous crossing. Intelligent Cross, for example, does periodic, stock-specific matches.[3] The goal is to reduce adverse selection by separating the crosses (randomly) in time, while keeping the interval sufficiently short to allow trades to execute in a timely manner (a trade-off that depends on the underlying stock being traded). Some venues offer an “at close” order that will be paired off against matching “at close” orders prior to the close, but executed at the official closing price after the close. While these orders essentially mimic orders routed directly into the exchanges’ closing auction (i.e., guaranteeing the closing price), they can provide explicit cost savings and/or superior functionality (e.g., ability to submit/cancel orders after the NYSE’s MOC cutoff time).
Market Data
Market data is an important input for dark pools, as the quality of that data determines the quality of the peg pricing. Significant delays in market data (relative to other feeds) expose pegged orders to “latency arbitrage,” as faster traders attempt to execute at the stale prices. To combat this, dark pools have tended to favor direct feeds, which have lower latency than the SIP and are often the same feeds used by the latency arbitrageurs themselves.
On one extreme, some dark pools use sophisticated techniques to minimize market data latency. For example, the pool may transmit data via microwave technology to reduce latency, switching to fiber automatically if weather is not conducive to microwave transmission. On the other end of the spectrum, some brokers still rely solely on the (slower) SIP feeds. While the SIP latency has been reduced substantially over time, direct feeds still have an advantage, as they are disseminated “directly” from the exchange instead of indirectly after having passed through the SIP first.[4] A common middle ground strategy is to use the direct feeds for the major venues, while relying on the SIP for smaller markets (e.g., NYSE Chicago). The intuition behind this approach is that the smaller markets rarely establish the NBBO on their own, so provide little meaningful incremental information.
Takeaways
This discussion highlights ways in which dark pools differ from one another. They also suggest various ways a buyside trader can use information in the Form ATS-N’s to enhance performance. First, because brokerage clients of some dark pool operators have advantages over others, e.g., higher priority, greater ability to choose which “tiers” to interact with, etc., such clients may benefit by biasing their trading toward the broker’s own dark pool. The key though is understanding the relative benefits provided when accessing the broker’s dark pool via the broker’s own algos/SOR and then measuring whether biasing flow toward that venue actually adds value – or just saves the broker some money.
Second, some dark pools provide clients with greater control over the counterparties with which they interact. By tailoring their interaction, say, to only “low toxicity” tiers, trading performance may improve due to reduced adverse selection (albeit at the risk of lower overall fill rates).
Third, because interaction rules differ significantly across pools, traders may be able to tailor their dark pool strategy based on the characteristics of the trade. For example, a trader working a sufficiently large order may benefit by utilizing a pool that gives large orders higher priority. Similarly, dark pools that provide price improvement to the resting order may provide an opportunity for a pricing advantage.
Fourth, some pools offer special orders tailored to specific benchmark strategies. A trajectory cross, for example, allows VWAP orders to pair off with one another, eliminating tracking error and potentially reducing market impact.
And lastly, traders may want to pay particular attention to adverse selection costs when accessing pools that still rely on the SIP for pricing. While latency differences have fallen over time, such differences have led to increased latency arbitrage and higher costs in the past.
In summary, the Form ATS-N’s provide a wealth of information. I highly recommend reading through these, as they are highly informative and could provide insights into how to improve trading performance (assuming, of course, you can suffer through the lack of formatting and bureaucratic lingo).
The author is the Founder and President of The Bacidore Group, LLC. For more information on how the Bacidore Group can help improve trading performance as well as measure that performance, please feel free to contact us at info@bacidore.com or via our webpage www.bacidore.com.
