Is Smart Routing Really Smart?
Not so according to recent research.
October 5, 2010
Last week, a former physicist announced his committee’s findings on the great flash crash of 2010. Before the report’s release, the New York Times said, “it will tell a clear story about what happened in the markets on that stomach-churning day, beyond simply pointing a finger at the perils of the kind of high-speed computer trading that dominates today’s markets.”
But high-speed trading is integral to the market these days – in some sense, it is the market. Whether it’s high-frequency traders seeking co-location arrangements to minimize latency so that orders don’t hang in time, or institutions trying to get their large orders worked in algorithmic blocks, it’s here to stay.
For most investors without direct access facilities to a market, that means using proprietary smart routers to get the best execution among multiple marketplaces.
Multiple marketplaces are not new in the U.S. They began as the SEC sought to short-circuit the seemingly cosy arrangements of Nasdaq market-makers taking advantage of the bid-ask spread. The resulting electronic communications networks sapped trading volume from Nasdaq. Their Canadian equivalent, alternative trading systems, have made inroads on the TSX’s former monopoly on share-trading.
All of which leads to smart routers, so that clients get best execution, regardless of the marketplace where it happens.
But how efficient is smart-order routing? It depends, suggests a recent Goldman Sachs paper.
“Over the past decade, markets have become increasingly fragmented as more volume is routed away from primary markets to alternative venues, like ECNs and dark pools. Traders must now choose not only the price and timing of orders, but also where to route their orders. Similarly, algorithms like VWAP [Volume-weighted average price] and TWAP [Time-weighted average price] that are designed to slice a large “parent” order into the market over time rely on smart routers to determine the best location to route each “child” order. Choosing the wrong market can lead to lost opportunities or increased costs, as orders go unfilled or need to be priced more aggressively to be executed.”
Not that “smart” routing is always that smart. Liquidity suppliers – i.e., sellers – often get rebates for orders that swim like fish in a barrel. Without them, there would be no markets. But the markets have since developed a “darker” side. You don’t know how many fish are in the barrel when you shoot in an unlit or “dark” pool where price indications are there, but not the size of the block.
Goldman Sachs takes two approaches to quantifying the impact of smart routing.
“For non-marketable orders [limit orders below/above the current bid/offer], we compare a relatively naïve strategy employed by one broker of choosing the market with the most advantageous rebates to more sophisticated strategies that attempt to maximize fill rates. For marketable orders, this naïve strategy is not permissible since marketable orders must be routed to the best-price quote in the market. For marketable orders, the naïve strategy used as a benchmark is to choose the lowest fee market among those quoting at the inside. The naïve broker therefore foregoes any opportunities to improve price, e.g., by “pinging” other venues for hidden liquidity at prices better than inside quote.”
One result would seem to be that maximizing rebates produces inferior execution. “We find that there are statistically significant differences in performance across broker smart routers for non-marketable limit orders. The naïve strategy employed by one of the broker smart routers in our study performs consistently worse than those of other brokers. The difference between the worst broker smart router and the best is about 10% of the bid-ask spread (0.4 basis points) for large cap stocks and 6% of the bid-ask spread (0.9 basis points) for small cap stocks. Even if buyside traders could capture the higher rebates earned by this naïve strategy, it does not compensate for the inferior execution prices. For some traders, this difference may seem insignificant. However, for cost sensitive traders (e.g., stat arb or other higher frequency strategies), this difference can be substantial. The performance across brokers was consistent across market capitalization, which suggests that some brokers are better at smart routing across all liquidity tiers.”
Still, when it comes to marketable orders (those above/below the current bid/offer) it’s a different story.
“For marketable orders, we find that there is generally little difference across smart routers. This is not surprising given Reg NMS price protections and that we require our brokers to execute only in lit market (i.e., no flexibility to utilize dark pools or cross internally).”
In the fast-moving world of electronic trading, there’s always the chance another analysis will be available … momentarily.