Dark Pools: Flash Orders, High Frequency Trading and Other Financial Market Issues (2009)

In finance, dark pools of liquidity (also referred to as dark liquidity or simply dark pools or black pools) is trading volume or liquidity that is not openly available to the public.[1] The bulk of these represent large trades by financial institutions that are offered away from public exchanges so that trades are anonymous. The fragmentation of financial trading venues and electronic trading has allowed dark pools to be created, and they are normally accessed through crossing networks or directly between market participants.
One of the main advantages for institutional investors in using dark pools is for buying or selling large blocks of securities without showing their hand to others and thus avoiding market impact as neither the size of the trade nor the identity are revealed until the trade is filled. However, it also means that some market participants are disadvantaged as they cannot see the trades before they are executed; prices are agreed upon by participants in the dark pools, so the market becomes no longer transparent.[2]
There are three major types of dark pools. The first type is independent companies set up to offer a unique differentiated basis for trading. The second type is broker-owned dark pools where clients of the broker interact, most commonly with other clients of the broker (possibly including its own proprietary traders) in conditions of anonymity. Finally, some public exchanges are creating their own dark pools to allow their clients the benefits of anonymity and non-display of orders while offering an exchange "infrastructure". Depending on the precise way in which a "dark" pool operates and interacts with other venues it may be considered, and indeed referred to by some vendors as a "grey" pool.

Whilst it is safe to say that trading on a dark venue will reduce market impact, it is very unlikely to reduce it to zero. In particular the liquidity that crosses when there is a transaction has to come from somewhere—and at least some of it is likely to come from the public market, as automated broker systems intercept market-bound orders and instead cross them with the buyer/seller. This disappearance of the opposite side liquidity as it trades with the buyer/seller and leaves the market will cause impact. In addition, the order will slow down the market movement in the direction favorable to the buyer/seller and speed it up in the unfavourable direction. The market impact of the hidden liquidity is greatest when all of the public liquidity has a chance to cross with the user and least when the user is able to cross with ONLY other hidden liquidity that is also not represented on the market. In other words, the user has a tradeoff: reduce the speed of execution by crossing with only dark liquidity or increase it and increase his market impact.

One potential problem with crossing networks is the so-called winner's curse. Fulfillment of an order implies that the seller actually had more liquidity behind their order than the buyer. If the seller was making many small orders across a long period of time, this would not be relevant. However, when large volumes are being traded, it can be assumed that the other side—being even larger—has the power to cause market impact and thus push the price against the buyer. Paradoxically, the fulfillment of a large order is actually an indicator that the buyer would have benefitted from not placing the order to begin with—he or she would have been better off waiting for the seller's market impact, and then purchasing at the new price.

http://en.wikipedia.org/wiki/Dark_pool