Commentary: Dispelling Dark Pool Myths

Don't fear the dark

Among buy side traders, there are misconceptions about how to evaluate and use dark pools. This article is an attempt to challenge some of these beliefs and offer some guidance on their use.  In my prior experience managing and analyzing dark pool performance, there is a big difference between the myths and the reality of dark liquidity.

1. Pinging is Bad
A ping, or a “blind ping”, defined as an Immediate or Cancel (IOC) order for a small quantity to find liquidity, has acquired a negative connotation.  Pinging has been cited by some articles as a sign of nefarious activity by high-frequency traders trying to sniff out liquidity and even to manipulate prices.  Actually many brokers’ algorithms use blind pings to find liquidity and increase the liquidity capture for their buy side clients. (These algorithms also attempt to find hidden orders on lit markets as well.) They do it, but they don’t talk about it too loudly because of common misperceptions. This leads to the second point:

2. You should prevent pinging with high minimum share constraints on your algorithms.
This applies only if you are not terribly interested in getting orders filled. If you curtail pinging you may eliminate the chance of crossing with some very respectable counterparties. The average execution size in most dark pools is only a few hundred shares. If you impose a minimum share constraint of 1,000 shares, for example, you would miss many crossing opportunities.

3. You can rely on studies to determine whether a pool is a good place for resting orders.
Actually, some studies may not have enough data to draw meaningful conclusions. You need hundreds of observations to determine trends and most orders resting in a dark pool don’t rest terribly long.  Even resting orders are usually of short duration, sometimes so short that they are essentially synonymous with IOC’s.  The orders that do rest for a long time may not be worth studying, as they might have very tight minimum share and/ or price constraints that make them trade too slowly (or not complete) often missing attractive liquidity.

4. The best way to compare dark pool liquidity is to compare crossing rates.                                                                          

 It is meaningless to compare dark pools by using crossing rates provided by different brokers. The problem is that while the numerator-the execution quantity-is known, the denominator-order quantity-is actually unclear. Brokers calculate it in different ways. Some clients put in new orders every time they change a limit; some use a cancel-replace in which many variations of an order are part of the same order. Does the broker count each new order, or try to link orders that just have limit changes? IOCs present another problem. If the pool has many IOCs, it may appear to have many orders and a very low fill rate compared to a pool with more resting orders. Liquidity analysis by a broker on their algorithms is somewhat better, but the fill rates are affected by how they send the orders and the sequence in which they send them.  For example, orders are often sent to the broker’s pool first and other large pools next so smaller pools get orders later, after most of the liquidity has been sopped up. Overall liquidity statistics from a third party often show a clearer picture.

5. You must avoid dark pools used by high frequency traders.
Again, ask yourself if you want dark liquidity. Some of the biggest pools have lots of high-frequency traders. And who are these traders? They include market makers, stat arb, and many other types of traders. It is not at all clear that you will be disadvantaged by trading with high-frequency firms. The issue is unclear as toxicity in dark pools can be very hard to measure or prove. What looks like a bad trade can be caused by many factors.

High implementation shortfall numbers may result from information leaking to other traders, or just normal adverse momentum in the stock.  A more subtle problem is a form of adverse selection caused by both the momentum in the stock and the order constraints.   Take a stock that has just had very positive news and the price starts to move away rapidly.  A trader might want this order done quickly, before the stock price moves further.  In this scenario, the dark pool is likely to have more buy orders than sell orders, and most orders would typically be for a few hundred shares. 

If the trader’s order is pegged to the mid-price of the bid-offer and has a high minimum fill constraint, it would be disadvantaged compared to other participants’ orders which are willing to pay the offer and/or trade in smaller size. The more constrained order may not get filled in the dark pool or get filled at a higher stock price than the other buy orders in the pool.  If you compare the execution to arrival price, it may appear that client had an expensive trade. The numbers do not show that the very measures the trader used to try to get a good fill proved costly instead.

Most practitioners use implementation shortfall and some variation of reversion (post-trade) as benchmarks to study performance, but so far, they have come out with very different conclusions.   Anecdotally, on one occasion when I was reviewing a dark pool, the algo brokers did far better than the high-frequency traders on an implementation shortfall basis and with no clear results on a reversion basis. (Please note that I did not examine unfilled shares due to the enormous volume of open orders.) 

On another occasion, in my observation of a dark pool it appeared that  one buy-side, long-only institution did far better than all other clients, including brokers – a result that probably resulted from stock selection alone.   Perhaps the answer is not to cut off trading with sources that have high-frequency trading, but to make sure the algorithm you use (or the underlying smart router) is clever about how to access that liquidity.

6. Traders use dark pools only for price improvement. 
Mid-point pegs are nice, but if you are looking to trade a stock with high adverse momentum, consider pricing aggressively.  Many traders cause their own adverse selection, missing opportunities to trade quickly when needed, in their attempts to capture spread savings. And don’t forget that another advantage of dark pools is potential size improvement by getting filled at a larger size than the quote. Many algorithms and smart routers send oversized IOC orders to seek attractive liquidity.

Please don’t misunderstand my argument: Dark pools are not without their risks. Neither are lit venues for that matter. However, dark pool providers want to ensure that investors come back and trade tomorrow.  This gives them an incentive to police their membership.

The bottom line is to put your trading objective first when using dark pools. If you have an order with favorable momentum or low urgency or both, trading slowly in a dark pool with higher minimum share constraints and mid-point or primary pegging might make sense. On the other hand, consider using a low minimum share value, accessing many pools and pricing aggressively when you really need to trade.

Marie Konstance is an executive director in electronic trading at Nomura Securities International. She was previously an algorithm strategist at a major broker-dealer and led TCA efforts at two other leading firms. She also has managed and analyzed dark pools in the past.  A graduate of Colgate University and the Harvard Business School, she is a Chartered Financial Analyst.   

The views expressed in this article are not necessarily the views of Nomura Securities International Inc.