Traders who make their reputations on relationships and negotiating skills are a dying breed. At least that's the conclusion of TowerGroup senior analyst Gavin Little-Gill, who recently tackled the complex subject of algorithmic trading. Little-Gill suggests that the trading desk of the future will prefer traders with the mathematical and programming skills to negotiate an increasingly electronic and quantitative marketplace.
Little-Gill expects the algorithmic trading of buyside order flow to double through 2006. Defined roughly as computer-managed trading that tracks a benchmark, algorithmic trading will grow 20 percent in 2005 and by 27 percent next year, according to a TowerGroup report by Little-Gill.
Most algorithmic trading is handled by brokers for clients, but an increasing amount is expected to be controlled by the buyside trader. The top three brokers in this area, as measured by share volume, are Credit Suisse First Boston, Morgan Stanley and Goldman Sachs, according to the report.
Commissions on algorithmic trades are roughly two cents per share, according to the report, about the same as for program (basket) trades. Those numbers compare to one cent for direct market access trades.
Little-Gill joined TowerGroup in 2001. The analyst spent nine years at money management house Scudder Stevens, where he built and supported the company's communications infrastructure. Little-Gill briefed Traders Magazine Technology Editor Peter Chapman on some of his findings.
Traders: Your report includes a pie chart outlining brokers' market shares, but you don't supply any numbers.
Little-Gill: That was deliberate. I have quite a bit of volume information. But it is all under non-disclosure agreements. I think I was actually the first person to be able to put out the volumes on a relative basis.
Traders: Isn't CSFB the leader?
Little-Gill: Everyone talks about CSFB being the leader. But there isn't much difference between them and Goldman and Morgan Stanley. There's a reason for that. The perception of CSFB as market leader has to do with its penetration of total trading desks. They are absolutely the leader on that count. Hands down. I don't think any broker dealer on that list would question that. They did a wonderful job in pairing up with Bloomberg very early on in the game.
Traders: But as far as share volume…
Little-Gill: The prime brokerage businesses – their hedge fund clients – of both Morgan Stanley and Goldman drive a tremendous amount of transaction volume. So it's a smaller client base, but…
Traders: Hedge funds trade more.
Little-Gill: Exactly. They have a higher propensity to use these algorithms than your standard long-only buyside shop. They have higher turnover.
Traders: The hedge funds using algorithms are generally the long-only variety? Not the strategy traders?
Little-Gill: Right. I don't see much use of algorithms on the back of strategies. Pairs trading, for example. It's all if-then statements as opposed to trading to particular benchmarks.
Traders: Commissions on algorithmic and program trades are about the same? Two cents per share?
Little-Gill: Pretty much. It depends on what you negotiate with your broker.
Traders: Yet you expect that number to drop. How come?
Little-Gill: Direct market access vendors are trying to move upstream. If a vendor can throw an algorithm in front of his DMA and charge 1.5 cents, he can increase his revenue by 50 percent. So, increased competition will come in the form of the DMA vendors pushing algorithms in front of their pipes. You are starting to see that today. Some of the DMA players think they can get a higher margin by putting in some more intelligence. The problem is you start to commoditize it.
Traders: You're talking about software vendors?
Little-Gill: Yes the Sonics, the Lavas. So let's say you are a big broker dealer offering an algorithm. You have a VWAP engine. But hey the vendor over here has a VWAP engine. Your VWAP is two cents. His is 1.25 cents. Is your VWAP three-quarters of a penny better? And more importantly, can you prove it?
Traders: DMA is basically smart order routing?
Little-Gill: DMA is anything that pushes an order directly to the marketplace under the nose of a broker dealer. You are sweeping marketplaces for liquidity.
Traders: Then how is smart order routing different from algorithmic trading? Recently, a direct access broker re-packaged its smart order routing tactics as algorithms. Are smart order routing and algorithms converging?
Little-Gill: Algorithms trade to a benchmark. With smart order routing you are intelligently pushing those trades into the marketplace. Pushing them so you limit the impact on the market. But you are not instructing them to trade to any benchmark such as VWAP or MOC (market-on-close).
Traders: Your report says the current crop of products is immature. Why? Because they mostly achieve average prices?
Little-Gill: We are still in the early stages. We haven't seen proof of differentiation yet.
Traders: How are brokers going to quantify the differentiation their algorithms offer? Transaction cost analysis?
Little-Gill: I don't know. The sellside will figure it out. In order to drive significant adoption, they need to figure out a way to quantify value.
Traders: Of their algorithms versus those of their competitors?
Little-Gill: Over the generalized activity. Average trades. VWAP is the most commonly used algorithm. It is easily understood. It ties directly to their best execution analysis applications. So they can measure it. It's easy to justify to a client. When you start to look at the other ones, the ones with which firms are trying to differentiate themselves, it becomes harder to rationalize them.
Traders: You don't feel transaction cost analysis is much good here because the benchmark is typically VWAP? So it doesn't work for any other algorithm?
Little-Gill: Right. If there is alpha information embedded in the trade, it doesn't get captured. So you are making a decision about how to trade a security based upon information over the course of the day. What happened before noon may be irrelevant if you placed that trade at 2:00 based on information that came out at 1:55.
Traders: Is there no difference between brokers' algorithms?
Little-Gill: Each of the broker dealers will tell you they have differentiated products. They have teams of PhDs rejiggering these mathematical models to provide that differentiation. So, clearly there is differentiation. But what it means from an execution standpoint is largely opaque to the buyside.
Traders: The hot new algorithm is known as implementation shortfall or arrival price. What is that all about? It has to do with the delay when an order passes from portfolio manager to trader?
Little-Gill: Well, there are many different delays that can occur that could cause implementation shortfall.
Traders: So what does the algorithm do for them?
Little-Gill: It depends on which broker's algorithm you are talking about. You have to look at each one. Most are looking at the arrival price of a particular trade and try to make the average execution equal to that arrival price.
Traders: Most algorithm trading of buyside order flow is conducted by the sellside? And not directly by the buyside desk.
Little-Gill: Correct. Only about seven percent of buyside directed trades are being pushed through algorithms.
Traders: Your report says that use of algorithms by large money managers is low.
Little-Gill: Yes
Traders: That's surprising.
Little-Gill: I spoke with seven of the top 10 buyside firms globally. Every single one of them said they thought they were behind their peers in their use of algorithms.
Traders: Why are they slow to adopt?
Little-Gill: They believe their trading desk is an advantage for them. That trading desks are profit centers. That they generate alpha through their trading activities. That is the philosophy. So using an algorithm means to accept mediocrity as far as they're concerned.
Traders: And that's because most algorithms such as VWAP and TWAP return average prices.
Little-Gill: Correct.
Traders: So they have no interest?
Little-Gill: Most are dabbling. They might use algorithms for certain portions of trades or certain types of portfolios. Nuisance trades, for example. The larger shops are dealing with much larger size orders allocated across multiple portfolios. They're dealing with different dynamics in terms of trades.
Traders: O.K.
Little-Gill: Also the largest buyside shops all use proprietary order management systems. The third party OMS vendors have integrated capabilities to provide direct access to these algorithms. But the large guys with proprietary OMSs have not done so. They must trade directly through the portals. That's less convenient.
Traders: Isn't the buyside confused by the sudden explosion in brokers' algorithmic offerings. What is the biggest problem? Which algorithm to use? Or which brokers' algorithms?
Little-Gill: Both. There's not much intelligence regarding who has the best algorithms. There is also fear of the unknown to some extent. And if you have seven brokers with six algorithms, that's 42 algorithms to choose from.
Traders: Thanks Gavin.