Robert W. Hogan Head of Client Trading Deutsche Securities Inc. Tokyo May 18, 2007

Robert Hogan is a Managing Director and the Head of Client Trading at Deutsche Securities in Tokyo . We have always known that the soft and mild demeanor of his exterior hid an exceptionally sharp mind and for a long time we have been trying to pin him down for a chat. Algorithmic trading, a subset of Quantitative trading, is somewhat of a niche and a somewhat esoteric area within the asset management and securities trading arena and you won't find your average, run of the mill CFAs or MBAs running the show here. Almost everyone in Robert's team is a computer scientist and a few are financial engineers. We recently had the opportunity of meeting with them and our Rahul Bhattacharya was finally able to sit down with Robert for a one–on-one talk. Here is a transcript of that conversation. Here is an excerpt from our conversation with him.

Team Latte :
What is algorithmic trading? How is this different from quantitative trading?

Robert Hogan
Algorithmic trading entails developing productivity tools for traders through software that helps automate the trading process . In algorithmic trading we write software to allow the trader to carry out the trading process in an efficient , productive way. Our first step is to try to understand the risk and cost associated with t rading at the micro level . Next we attempt to assess the impact of the overall market at the macro level and then design better trading techniques that reflect these insights . Ultimately the goal is an optimal trading process, balancing off expected market impact with the risk incurred by extending a trade over a longer time horizon.
There is an overlap between algorithmic trading and quantitative trading. At a general level, quantitative trading addresses the question ' what do we want to trade and what would be our investment decisions', with algorithmic trading then addressing the question of ' once we have decided on what we are going to trade how do we implement those trades in an efficient manner'.Considered separately, algorithmic trading focuses on the provision of tools for agency trading - tools which are designed to provide our clients with effective and efficient execution technology . Quantitative trading on the other hand is systematic, process oriented trading whereby one develops measures of value that attempt to capture the future returns of a security. These measures of value are then used to direct a systematic trading process. Algorithmic trading is more about execution strategies, dealing with questions like how a trader should trade a portfolio in the market balancing risk and the expected implementation costs. Really though this separation between portfolio construction and optimal execution strategy is an artificial one.
Ultimately we need to integrate the portfolio construction process and the trading process so that expected alphas and expected transaction costs all feed the investment process

Team Latte:
How has the landscape of algorithmic trading changed in the recent years?

Robert Hogan :
In the last couple of years there has been a big shift in the dynamics between the buy side and the sell side, especially in the U.S. and Europe . Professionals on the buy side are clamoring for more sophisticated and accurate tools and systems to estimate trading costs and facilitate execution decision s. Meanwhile professionals on the sell side have been moving up the learning curve in terms of tools and systems for efficient and effective execution of trades. Overall volumes have grown tremendously and the trading process has become quite complicated. This has created a slight disconnect between the buy side and the sell side. We are trying to rectify that imbalance by empowering our customers - the traders and fund managers on the buy side - with better tools and technologies from our repertoire so that we can help bring them up the learning curve fast er on optimal execution . Japan and the rest of Asia are a little behind the other regions but catching up very quickly.

Team Latte:
There are two major concerns while trading emerging market or small cap equities. One is that of volatility and the other is that of liquidity. One may have an impact on the other. How do you account for these two parameters in your execution decisions?

Robert Hogan :
This is an interesting and a fundamental question. When you place a bid on a market when there are no other bids this is an act of market making. You are effectively supplying liquidity to the market. What matters here is what kind of risk premium the trader is willing to demand for the trades. Liquidity or the lack of it will clearly impact the underlying volatility of the security and the problem is closely related to that of risk pricing.
In other words illiquidity, volatility and expected market impact are all linked together in a rational way and a careful approach to algorithmic trading will apply this kind of reasoning in the trading process.

Team Latte:
A key part of managing a quantitative trading program, such as a statistical arbitrage strategy, is working with detailed software programs that try to capture market anomalies. How much is the technology/software input in your strategies? We believe that your team is primarily composed of Computer Scientists. Are they engaged in writing software algorithms to spot price anomalies or capture other market imperfections?

Robert Hogan :
Yes, you are right . We do have quite a few bright computer scientists and financial engineers with formal degrees working for us. In the field of a lgorithmic trading a broad range of skills are required. But chief amongst these skills are financial engineering, software design and software engineering. When we write algorithms, we need to model market impact and the risk of holding equity portfolios and then transform that information into efficient trading techniques. We need to work with large volumes of data around market depth, trading histories of multiple stocks and handle very high transaction counts to assess intraday patterns around volume and liquidity . The skills to do this ultimately cut across both financial engineering and software design with software engineering skills becoming a requirement. We have client s who trade over 10,000 orders every day and this raises issues of scaling and capacity - software systems need to be developed in a manner that can handle this.

Team Latte:
A major concern that all equity traders face is that of the quality of data in a correlation or a covariance matrix. When there are large number of stocks, including small cap and some illiquid stocks, in a trading universe the quality and functionality of a correlation or a covariance matrix becomes critical. How do you resolve that issue within your team?

Robert Hogan :
Once again this is a pertinent question. Today we have a long history of techniques to estimate correlation and covariance of stocks. We take empirical and backward looking data to estimate a covariance or a correlation matrix using techniques such as Principal Components Analysis. We need to back test and run the data on established models. A simple model can have an awful lot of explanatory power , and the input data, such as correlation and covariance, should be tested with such models. In short there is a rich field in finance around modeling equity risk properly. The techniques developed in this field are precisely what are required to avoid the pitfalls and shortcomings of a historical full covariance matrix approach.

Team Latte:
It would be nice if you can tell us a bit more about some of the "light weight" algorithms that your team has recently developed to help your clients manage their portfolio better.

Robert Hogan :
Some of the "light weight" algorithms that we have designed are "pegged iceberg order s" and "I would" parameters . Pegged iceberg orders are really very simple. You specify your overall quantity, what quantity you want to display on the book and how to price it versus the bid or offer. Let us consider a case of a buy order, pegged to the bid, which might be appropriate for an illiquid name. While you specify what size you want to display on the book, we apply a modest randomizing adjustment to prevent predictability. The algorithm places a buy for this quantity on exchange at the current bid. As this exchange order is filled, another new order, once again at the display size subject to randomization, is placed at the then prevailing bid. Very simple but it replaces an activity human dealers find tedious but necessary for illiquid names.
The "I would" parameters are really more of a "mix in" for an algorithm than an algorithm in its own right. In other words "would parameters" get used in conjunction with a more standard strategy like implementation shortfall or volume weighted average price. The "I would" parameters allow the trader to state a level where, if the market moves favorably, in other words in their direction, they would accelerate the execution process. This level can get expressed either as an absolute price or a level relative to an index. This type of intuitive flexibility means that our clients can customize the way in which our standard algorithms behave in certain conditions. This approach is one we consider preferable to rolling out a large number of less-flexible algorithms that differ only slightly in behavior
For example if you were buying 5401 Nippon Steel, full quantity of 2 million shares, you might say "I would" buy 100,000 shares when the name underperforms the TSE Iron and Steel index by one half per cent relative to its arrival price. This is actually common trading room terminology in the U.S. but the "would parameters" let you express this in an algorithmic trade.

Any comments and queries can be sent through our web-based form.

More Asia's Brightest >>

back to top