One of the themes we have been active in discoursing over the past few years involves the conflicts of interests in order routing. Our discourse has focused on the wisdom of the “maker-taker” exchange pricing models, the nuts and bolts of algo order routing, as well as the plumbing of smart order routing. In general one has to question at least the wisdom of increasing the opportunities of increasing numbers of destinations and market players “touching” order flow as it makes its way to public and displayed markets.
This morning we would like to direct your attention to an article on the Tabb Forum written by Chris Sparrow, called The Teenage Order. He discusses TCA, and its value as it evaluates the parent order slicing off child orders that are sent to be executed. He makes some great points. In considering a VWAP algo, where the parent order is the large order in its entirety, and the intraday slices are the child orders, he notes that this works well in a single venue execution environment (think centralized exchange- which is a notion of yesteryear. He goes on to say, however, that in our modern multiple market execution environments, such simple TCA is lacking. It is lacking because the VWAP algo no longer sends child orders to a single venue, but rather to a smart order router (SOR), which in turn sends its own child orders to a myriad execution venues.
We diagram that process to look like this:
Sparrow goes on to acknowledge the dangers of being subject to “latency arbitrage”, a topic we have covered in the past:
For example, if we want to measure the ability of a router to capture visible liquidity, then it is important to distinguish between situations in which one venue is accessed versus situations where several venues are simultaneously accessed. The algo (strategy) may determine that it needs to buy 1,000 shares actively and issue a marketable order that it sends to a router. If there are more than 1,000 shares available at a single venue at the best price, the router may send a single child order for 1,000 shares to that venue and avoid latency arbitrage.
Alternatively, if there are five venues each with only 200 shares at the best price, then the router may send five child orders, one to each of the venues displaying the best price. An analysis of router performance should treat these two scenarios separately.
There is a much higher probability of being subject to latency arb and missing shares in the second scenario than in the first.
The article is a good one, as it lays out why we all need to be thinking about how our orders touch the market place. The architects of our modern markets have made it needlessly complex, not to mention treacherous for each of us, for their benefit and not ours.
Understanding each of our specific processes is the first step to correcting them to avoid the sweaty handshakes we desperately wish to avoid.