As the heat rises in global securities markets, blocks have melted into streams: trades that were previously executed as one big block of shares are being broken up into a series of smaller trades and reported to a customer at an average price. Among the most visible evidence of this trend are the precipitous drops in share quantities per trade, down from 2300 to around 300 on the NYSE.
Of course one of the drivers behind this trend is the increased use of algorithmic trading systems for tactical execution. Many brokers offer to “rent” their algorithms for use by their institutional customers (and, increasingly, their retail customers), which presents a crucially important question: what constitutes “best execution” in the world of streams?
Most recently, best execution requirements were re-framed by Reg NMS, specifically the regulations that require brokers to determine the best trading venue to execute an order based on a subjective mix of price, likelihood of execution, and price improvement. The regulation is only written, however, to protect orders that exist at one instant in time. In moving from block to stream, a new variable has been introduced – in addition to where to send the order, now you must decide when to send the order.
Of course, with its roots in common law of agency, the onus of best execution has been on brokers since long before Reg NMS. In fact, the closest analogy for today’s broker-supplied orders is “not-held” block orders. Unfortunately, as customers of Knight trading found out, the lack of transparency in those orders can come back to bite them in the assets.
Customers and investors are starting to understand that even in the absence of order front-running, getting a handle on the costs of using outsourced algorithms is difficult. Increasingly, they’re are turning to software to perform the transaction cost analysis of these algos (more info in Daniel Safarik’s excellent article What is Best Execution?). But, at the end of the day, it’s a science project if you are not privy to the inner workings of the algorithm.
An increasingly attractive alternative is to use tools like the Marketcetera Platform to encode your own trading rules in software and use DMA connections to construct your own tactical trading applications. You get unprecedented transparency (it’s open-source after all), which makes trade impact analysis more straightforward. And more importantly, you can be your own arbiter of best execution.