PDQ Launches 20-Millisecond Auction System for Equities
November 9, 2011
PDQ Enterprises began using an auction system for trading stocks that works in 20 thousandths of a second.
The technology developer said it began using the system on its PDQ alternate trading system today (Wednesday, Nov. 9, 2011).
Under the system, an institution can place a buy or sell order on a block of stock. The system does not publish the name of the firm or the quantity.
Instead, it asks algorithmic trading systems to make their best offers on a given stock symbol, such as IBM.
The “electronic algorithmic crowd” then, in effect, competes for the order. The system then fills the order, according to parameters on acceptable prices and quantities set by the institution.
The system collects responses, builds a book of counter orders for 20 milliseconds, then matches responses in order of price, time and quantity.
The approach is based on a patent received in 2008 by PDQ founder Christopher Keith, a former chief technology officer of the New York Stock Exchange.
That patent is not for an auction system. Instead it covers an electronic procedure that specifies trigger conditions and what to do when the trigger conditions occur, or do not occur, within a specified time.
Broadly speaking, this “ corresponds to an order handling strategy used by a trader,’’ according to an abstract of the patent.
The auction is instead an application of the sort of procedure set out by the patent. And what PDQ hopes will set its approach to equities trading from that of other dark pools.
"We of course have fancy dreams that it will be great, but certainly it's just a matter of whether the marketplace decides this is the methodology it thinks it wants to use,’’ said PDQ chief executive Keith Ross.
"Given the hum of the industry that people think they're getting taken advantage of by the high-frequency group,” Ross said. “Our thought is that this a way to aggregate liquidity around the order and build the average order size again."
The system is efficient and confidential, protecting anonymity of institutions and precluding leakage of details such as quantity involving a trade, according to Ross.
The algorithms that define what traders are willing to pay and for what amount of shares in a given symbol are, for instance, stored on PDQ servers. Even if the quantity is disclosed to the algorithms, that information is shared with all trading firms’ algorithms as the same time. And the information does not get sent to the firms themselves.
The information remains “contained” on PDQ’s system.
Ross said the “small pause in the trading of an order” that allows the auction to take place “can create a more efficient market structure.”
PDQ currently matches between 20 million and 30 million shares of stock in average daily trading. Until now, it has relied on a “first to respond” system.
PDQ will allow its customers to continue to use the “first to respond” system and “opt in” to the auction system, when they’re ready to use it, Ross said.
