TST Interviews Nunzio Tartaglia: Quantitative Trading And Technology

In 1989, Nunzio Tartaglia parted company with Morgan Stanley & Co., where he made his name as the man behind the pairs-trading strategies of the firm's Advanced Proprietary Trading group. Tartaglia has now emerged with his own research and trading operation. TST asked him about life, the Universe and everything...

Tartaglia: Well, I have been looking around to see what the various opportunities were both in large corporations and in smaller situations, as well as going into a very small shop. The conclusion I came to is that I could no longer take any corporate environment. It's anathema to technology, to research, to advancing ideas. It doesn't lend itself well to creativity. So you have to go to a smaller shop, and then the question became how small? What kind of resources do you need? And that was the critical question in terms of size.

What I realized that there were a lot of software products in development, that the machines were getting very fast and very inexpensive and that in fact we could go to a very small shop and have a lot of fire power. So, that's what I decided to do, I joined Rich Bartels. Rich came out of Barr Rosenberg Investment Management, where he built the technology trading infrastructure for that firm. He was in Boston and we combined and set up an office here in Lexington. We just formed a joint corporation called DerivaTech Inc.

Tartaglia: I've known him in a quantitative area for 10, 12 years. Most of the quantitative people around the country I have known at some time or another.

Tartaglia: If I had to put my finger on it in one statement, with the kind of things I was trying to do, I didn't really belong in that kind of organization. There were always conflicts and cross-currents and politics and bureaucracy, etc. For example, the whole group doesn't exist any longer. So, that'll give you an idea of how much backing they gave to us.

In essence, while we were making money, they were committed to us. And when the profits started to drop and we had to contract, they realized that we're investment bankers, we don't need this business.

We never failed to make trading profits. These are general ideas, they don't apply just to Morgan Stanley, they apply to a corporate environment. For example, in Morgan Stanley you couldn't make $23 million like the trader did at Salomon Brothers because he had an arrangement whereby his income was tied to his profitability. In a corporate environment it doesn't matter how much money you make, they'll only going to give you a certain amount. So what does that mean? That means in one year I delivered like $30 million in profit, and we didn't get paid anywhere nearly commensurate with that.

So, what does the corporate environment force you to do? It forces you to build infrastructure. Because you say if I'm not going to get paid this way, I might as well get paid in expanding my operation, bringing in more people, bringing in more technology, doing more things: in other words, building overhead -- something that I never did and never wanted to do. But, being in the environment that I was in, I said this is the only thing they allow you to do, given the corporate structure. So that's what I did. I then built the organization up to a maximum of like 53 people and had a $20 million expense. Well, what happened was the profits fell below that. Then I had to contract. Then in the contraction process, that's when things got messy and hairy.

The essence of technology is that if you know how to use it, you can be highly productive and don't need all the overhead. Once you hit a corporate environment you're a salaried individual with a bonus tied to your profitability. Well, I said to the guy I was reporting to in the year we had our big numbers, in August I said to him: "You know I can go to the beach here for the rest of the year. I'm not gonna do anything if I don't get paid any more. What's the point of my bringing in any more money?" And that becomes the corporate problem. And that's why in a sense it kills creativity. It kills aggressiveness, it kills a desire to break into new ideas and new territory and introducing new technology. It doesn't reward that.

Tartaglia: We think it is a new idea in that DerivaTech's approach to trading is really a trading laboratory. The idea is this that a pension plan may not have the wherewithal to involve itself in technology and the learning of technology and the development of trading strategies. They in general don't have that for the amounts of money we're talking about: $10, $20, $30 million, small amounts of money.

In a trading account we will develop the trading strategies and the technology. There is a fee that basically serves as a consulting fee on the running of that money. We will participate in the profits that it makes. But the essence of the relationship is that we're allowed to continue to research and look into new technologies, new software packages and development of new trading strategies that we will share with them on a consulting basis and help them to apply to their broader investment strategy.

We'd be trading with their money on their behalf.

So it's basically a small amount of money, R&D money is what it is. What I enjoy most is the research laboratory environment. It's what we had in Morgan Stanley, it's almost like I got cursed for one of the comments I made. You know, when you're developing a strategy and it's making money it's almost boring. Because there is nothing to research.

It's when things go wrong that the challenge begins. I'm a research individual, I've always loved research and the challenge of research and this allows me to develop a small group here with Rich where we can constantly be researching new ideas and technology and applying them to the trading environment.

Tartaglia: I've been here about six months. We have an off- shore client and we have three or four very big recognizable names that we're talking to, pension plan sponsors.

We're only going to be doing this with four or five names at the most. We are not interested in building an asset base worth billions of dollars, I mean probably at most we'll manage about $100 million in trading dollars. It becomes too unwieldy at that point. We believe that if the strategies work and there are high returns with a small number of people, that is really all we need. So it's not an empire- building effort.

We're basically working with Jefferies & Co. The main products that we are working with are Integrated Analytics Corp.'s MarketMind, and they combine that analytic software with the Jefferies trade execution system called Quantex.

We are the major beta test site in that effort. I'd seen just about everything out there and the Quantex product to me was going to be the most powerful. Now again, it's not a thing that you walk into a trader with, you have to know what you're doing. But for people like myself, who want to use it in a creative way, it always has shortcomings, it doesn't do enough.

Tartaglia: We looked at a lot of products out there. My key interest was that I didn't want to have to do programming. For example, MarketVision has very powerful graphics and you can put strategies in it too. But that means I have to program and I wanted to minimize that.

I was heavily involved with Frontier Financial Corp. It's another approach. I've spent a lot of time working with neural networks, adaptive pattern recognition and all these fancy phrases. The biggest problem I have with all of that is that it's a statistical approach which is attempted to be used in a dynamic way. So, as things change you're reading the new pattern, you're changing as well. I think the predictive value of these approaches has not been established.

For example, you can never tell when you're wrong, why you're wrong. There is no causal analysis that comes out of that approach. And again, being a physicist, I can't handle that, you know that too much in the world of finance is based on statistics. Well, a non-stationary, non-linear world makes all those statistics worthless.

It may work for you at a time and it certainly will work for you over the period that you developed your models. Otherwise you wouldn't have the model you have. But out of samples those programs tend not to hold up.

What Frontier's Ed Bosarge does is, he lines up a bunch of models and asks them to compete with one another. And out comes the top model that has been able to do the best predicting over the period he analyzed. What happens is, every day there's a new best model.

Integrated Analytics allows you, with a mouse and windows, to pop up and put commands and create programs without having to go through the elaborate effort of designing programs to handle every problem. So you can do some very powerful things. For example, the way they handled data, the whole database management effort is totally taken care of. People don't realize how critical the whole data side is, and the data management.

We use ComStock data. If I create a program which tells MarketMind to execute a certain strategy, the system will automatically dial out, it will look over the program, look to see what data is needed; it will automatically dial out to Knight-Ridder, download the data, prepare it and bring it up on the screen and start tracking it in real time.

I don't have to worry about the database managing side at all, and that is the most critical thing that you get involved in whenever you're trying to develop a strategy.

I have all the rules right in front of me, all the values, and it's all done in real time. Plus, I can manage baskets as if they were single entities. So, for example, our whole basis for trading here is arbitrage. And we're doing what I call, some people are calling it synthetic instruments.

I prefer to call it security emulation.

Tartaglia: Yeah, I've seen the Algorithmics stuff. I don't think that's the road I want. For example, the creation of instruments via the options model to me is not the way to go. I have always detested the options models, I believe they're a debilitating use of mathematics, because it'll focus you on a statistical approach to analysis rather than relationship that you're involved in.

The point is that through the emulation, I can create a security which is a basket of instruments. And as long as I can find another basket that emulates it with different instruments then I can trade them in an arbitrage.

Tartaglia: You can have a combination. Let's say for example, you have long rates on one side and short rates on the other side, Well, could you combine that with some equities? Long bonds with certain equities that'll respond to short rates vs. short bonds and equities which respond to long rates. And in that way you have different instruments, but statistically you've created two baskets that look alike and so you can arbitrage them.

So that's the essence and foundation of what we are trying to do. You can do this across a broad range of instruments, and the approach we're using is to create algorithms which don't require you to work with any particular instruments. For example, you can throw in London stocks, Tokyo stocks; whatever you can trade electronically and quickly you can include as instruments and possibly to build baskets around.

Tartaglia: We are trying to work with a brokerage firm in London. What we are trying to do there is have them take our baskets, which they know are hedged baskets, at the end of the day, closing prices. So we just get a set price. That gives them merchandise, they are non-informational trades, we then buy them back when we're ready to buy them back.

Tartaglia: What we are trying to do is to play strictly where we can do it electronically, like the futures market. Here we have to talk to people on the floor. We use the futures brokers and at times we need information, so we have what I call information brokers. We need people to chase down information. So we do have to use, unfortunately, non- electronic means to get trading done yet.

Tartaglia: Oh, absolutely. We're just salivating, waiting for it to get going. But what's clear is that all of the entire world ultimately, I guess five, ten years from now, is going international and going derivative instruments. Pension funds don't know how to use derivative instruments. They don't understand them. Yet, a farmer out in the plains of Iowa knows enough how to hedge his wheat fields. And what is a pension manager's wheat? His ownership in stock and bonds. That's equivalent to the inventory of a farmer with his field. And yet a farmer knows how to hedge his risks using derivative instruments and security managers don't.

Tartaglia: At the banks, they generally have people who are specialists in currency. The banks deal in currencies all the time, they have the currency desk and so they're pretty good at trading the currencies. But from our prospective part of the problem you have stems from the broker's side. With the bank side at least their desk is dedicated to what it does.

What I attempted to do in creating an analytical proprietary trading department in Morgan Stanley was to cross all those bounds. I didn't need to be specialized in anything, I could use any and all instruments, create a portfolio approach to trading. So, that's basically where I am right now.

The advantage the brokers had was the cost was significantly reduced, but unfortunately the Morgan Stanleys of the world didn't have the same vision; and Morgan Stanley is among the more advanced of the brokerage firms. The time had not come for them to do it, but my feeling was to just be where you are in your operation and attempt to make people understand that you're concerned as their leading scout. I scout the future for them. To come back and be a consultant for them, the way things are going, they can then move at their own pace and then have some comfort in knowing that they can be dealing and working with us at the forefront.

Tartaglia: I think we head back; you go through these periods. I've been around the Street long enough -- about 18- 20 years now -- and I've seen the bull markets and the bear markets, and I'm still one of those people that remembers intimately the '82 bear market. You just go through these periods; they occur and they reoccur. And the same questions occur when things are bad. Everybody projects to infinity; it's never going to end, it looks like a depression. One of the most predicted depressions in the world is occurring right now. And I even read it in the Boston Globe.

Tartaglia: The debt of the '80s set up the '90s to be the equity decade. You just built up debt in the '80s and now the capital-raising function has to be moved back to raising equity. What do you think is happening with the banks? They're being asked to increase their equity. Now all of a sudden what stocks are in favor? Stocks where you have manageable debt with a reasonably high equity. Overleverage is out.

Tartaglia: The workstation is the single most critical thing. The fact that you have an open software environment, an open hardware environment, where you UNIX, C and Fortran... We have a Sun SPARCstation-1+. Part of the algorithms that we're developing run sometimes a couple of days. We had one program on for 48 hours. Now I need a little bigger machine, so we are in the process of getting a SPARC-2, which will give us a little more speed, and if I get the run down to 12 hours, that's fine. Because these are free machines you don't need a machine room. They sit there and you go home to bed and you come back the next day and the program is run. So even a 12- hour turnaround is not bad. 48 is a little painful but you can do that over a weekend.

We have three Sun SPARCstations: one is to monitor the futures market in real time, one is to monitor our equity baskets in real time and also services the trade execution via Quantex.

We have another Sun, which is also a Quantex workstation, that my colleague and I work on as a research and development machine.

Then we have 386 PCs where we download, for example, our trade executions that we've done during the day onto a Lotus spreadsheet, where we can create our reports at the end of the day. We have another machine which puts up some additional algorithms with additional feeds: ComStock and [Infotechnology Inc.'s] Signal for backup. So, we have some quote pages up in front of us as well as on the PC.

There are four people here: Mark Gimble and I are the research development/strategy development people and Rich Bartels and Steve Killian are the traders. It's this kind of a thing that I tried to do at Morgan Stanley. You can't separate trading and the knowledge of trading from the technology and strategy development. We can develop perfect baskets and then I have to give them to Rich and he'll have to tell me how these trade. How can you trade them? What kind of strategy can you employ to make them make money? Or what problems you run into in trading that we can go back, research and create a better software for monitoring and trading these things.

Tartaglia: Yeah, it's a constant dialogue. It's not just a research room that's developed an idea and passes it to the trading room and they make money.