Greg Tusar, head of the Goldman Sachs Electronic Trading business in the Americas, sat with Traders Magazine recently to discuss a number of issues affecting equity trading today. Among the topics addressed were developments in market structure, as well as how algorithms and dark pools are evolving. An industry veteran of 18 years, Tusar offered some insights on the rapidly evolving marketplace.
He began by speaking about how order handling and routing-a process not easily understood by clients-has gotten far more complex over the past decade. The solution clients have been requesting is for brokers to make both processes more transparent to institutions, so they can make more informed choices about where to send their orders. Transparency, which Tusar supports, would give clients more confidence in how their orders are handled and where they’re routed, he added.
Traders Magazine: Is there skepticism these days from money managers about where their orders are getting done?
Greg Tusar: I think there’s a strong desire to understand order routing practices better. I think there’s also a desire to understand how order routing practices relate to algorithmic performance.
TM: Could you provide an example of the need to understand the process better?
Tusar: I think order routing practices can be made more understandable and transparent through two primary means. First, by providing clients with reports and real-time information as to where their orders went. Second, with modifications to Rule 606. Rule 606 could be modified to include all of the venues in which a broker searched for liquidity in the process of executing an order. While not client specific, it would generally give clients a sense of the order routing methodology used and the number and nature of venues where orders are executed.
TM: Just like the ICI letter said, they want to know where their order went and where it didn’t get executed.
Tusar: Absolutely. We currently provide the venue of execution back to end clients both via FIX and into our REDIPlus front end software. However, if a client is also interested in how and where we sought liquidity, and not just where the order was executed, we can provide that information through separate reports and files.
TM: But clients still have questions?
Tusar: One questions we get is: How do you simultaneously manage fees in the maker-taker world and the optimal placement of limit orders? In addition to the two ways I previously mentioned we can increase transparency, what we proposed in our recent Concept Release response is passing rebates and removal fees back to the end client. This could happen by making the execution price be inclusive of rebates or fees. In the case of retail orders, that would mean that the end customers are the beneficiaries of payment as it would appear in the form of price improvement.
In the case of other clients, choice of venues for limit order placement would be driven by their desire for either higher fulfillment or rebate, allowing the client to decide which constitutes best-ex from their perspective and the choice is reflected in their execution price.
TM: Switching gears, you mentioned the customization aspect of algorithms. We recently wrote about a large manager who basically profiles the portfolio manager. Is that a new trend? Is it something you expect to see more of?
Tusar: I think it’s a trend. It’s the next level of customization. Many clients have asked for help in making the algorithm selection process more scientific, mainly by working together to profile by a variety of characteristics: urgency, size, market cap, etc. In working together to do that, we in effect customize an algo which may reflect orders from different funds, PMs, traders, etc.
TM: So, would that be the next generation of algorithmic trading development? It’s going more on the data from past trading decisions made by people in an organization?
Tusar: That’s right — working together to look at prior experience to help inform future choice of algorithms. The other thing we’ve spent a good deal of time on is getting back to the basics of smart routing, evaluating optimal places for non-marketable limit orders, and creating a next-generation framework to evaluate liquidity pools overall.
TM: What’s the thought process?
Tusar: Things aren’t static, so you need to be constantly re-evaluating routing decisions and their impact on overall algo execution quality. That’s with respect to both venue selection as well as use of order types like hidden "peg-mid" orders. Inverted pricing models at exchanges are an example of what may change how and where others search for liquidity, which may have substantial impact of fulfillment.
TM: It’s almost as if the algorithm and the routing are really the same … a router is a router.
Tusar: It is, but in such a competitive environment it requires a constant commitment to improving infrastructure. For example, being able to react in a world that requires microsecond speeds in order to carry out client instructions. As an example, we are starting to deploy hardware acceleration to make sure we have the most timely data possible.
TM: You mention doing more block contingent orders in your dark pool, Sigma X. Can you expand upon that?
Tusar: Essentially we allow clients to simultaneously search for flow in multiple pools, yet stand ready to consummate a block should the opportunity arise. For instance, a client’s total algo order is for 10,000 shares. Our algo will be working small portions of that order, say 500 shares, at several different pools. Meanwhile, they can have a block contingent order for the total quantity in our pool that will only execute with another block. In other words, the client avoids over-execution risk and automates what they may have done manually, in a negotiating system.
There’s clearly a desire to increase the print sizes that people are experiencing, instead of trying to re-assemble the blocks. We think this will get clients there without requiring a change to their workflow.
TM: So, it’s just an ordinary cross through a dark pool, with the ability to not over-execute?
Tusar: It’s more about managing a large order, across multiple venues, in an intelligent way. We’re allowing clients to say, "Work my order algorithmically in all liquidity centers, but if the opportunity to cross with a natural arises, then get it done." Historically clients have had to choose whether they want to commit their order to an algo, or try to negotiate a block in a dark pool. Our intention is to provide them with both options simultaneously, and all they have to do is check a box.
TM: And this process takes how long?
Tusar: It’s fast … measured in milliseconds.
TM: You mentioned the circuit breakers implemented after May 6 might need calibration. What are your thoughts about limit-up/limit-down?
Tusar: The benefit of limit-up/limit-down is that it doesn’t actually halt trading. It allows price discovery to continue. Buyers and sellers will continue to meet and trade, just not at prices that are below a certain level. So you have the effect of having a pause, which seems like the right strategy. But you have the added benefit of trading at least being continuous. That’s particularly why we thought limit-up/limit-down is a better approach.
TM: Are there any issues with limit-up/limit-down?
Tusar: Well, coordinating limit-up/limit-down systemically might be technically difficult to do. So, in terms of there being a quick response, the current solution made sense. It might need further calibration to determine whether, say, down 10 percent over a five-minute window is the right level? Does it need a different mechanism? At least it’s a step in the right direction. The other thing we point out in the Concept Release is the concept of the consolidated audit trail, which for a host of reasons related to the industry’s ability to really understand the dynamics of market participants–that they have more rigorous (both regulatory and academic) study of the impact of the changes of the last decade is critically important.
And the idea of having a database that has systemic market activity in it that would allow you to do those kinds of studies we think should be prioritized. We featured that in our letter, also.
TM: What about the issue of speed? You’re saying that you have to be faster all the time?
Tusar: Yes, it’s one of these things where the journey is more important than the destination. Market participants have to be constantly committed to finding the next place to be more efficient, be better at acquiring the market data faster … all in the name of better executions for algo- and smart-routed orders. But we’re in a marketplace where everyone is competing on technology infrastructure, so we’re as engaged as anybody in infrastructure improvement. It’s very important. For example, if we have a client order who wants to buy 100,000 shares of a stock. When an offer shows up that’s at a price that our algo determines is attractive, we have to be as fast as anyone else out there to react to that quote and to lift that offer.
And so there’s really no such thing as "fast enough" when you’re executing client orders; we have to be competitive.
TM: This whole issue of high-frequency trading and speed … the concern is that HFTs can sniff out institutional orders and run ahead of them. Obviously, you don’t want that to happen.
Tusar: I think the key is to have well-designed order-placement strategies. And it goes back to our research on smart-order routing techniques that simply don’t leave the breadcrumb trail that allows for that to happen.
TM: It’s all about the routing strategy, then?
Tusar: It’s about the routing strategy, and the order placement strategy. The best techniques and the best mitigants to prevent what you just described from happening are pretty simple blocking and tackling kinds of things. In other words, randomize your order size and placement of orders, using "Minimum Execution Quantity," because I don’t want people pinging me for odd lots and then finding out I’m there and stepping in front of my order.
(c) 2010 Traders Magazine and SourceMedia, Inc. All Rights Reserved.
http://www.tradersmagazine.com http://www.sourcemedia.com/
