A San Francisco start-up is betting that hedge funds and prop desks need a faster way to bring their trading strategies to market. The company, known as 4th Story, has built a trio of servers that find, test and operate trading strategies across all asset classes. Founded a little over a year ago by two technology veterans, the vendor is achieving some success in the U.S. and abroad. The clients so far include Sofaer Global Research, one of the world's oldest hedge funds, and investment bank Barclays Capital.
The technology reduces the time and hassle of building strategies by traditional means, according to 4th Story. It relieves traders of the need to install the "plumbing" every time they design a new strategy, freeing them to focus on the strategy itself.
Currently, the most popular software used by traders to write strategies is Excel by Microsoft and Matlab by Mathworks. The use of Excel is so ubiquitous that an entire industry of "add-ins" has formed around it.
4th Story is not the only player offering strategy testing and operating software. At the institutional level, hedge funds can also turn to vendor Neovest, for instance. At the retail level, daytraders can access software from broker TradeStation.
4th Story, which employs six people, was co-founded by CEO Steve Smith, an IT veteran with Hambrecht and Quist and Charles Schwab & Co. Co-founder Philip Clem, formerly with vendor Longview International, is chief technology officer.
Traders Magazine Technology Editor Peter Chapman spoke with Smith about his firm's technology.
Traders: Your first product was 4S.Yellowstone. What does it do? Why bring it to market?
Smith: We were trying to build an open-ended platform for arbitrarily complex trading strategies. Order content' or alpha strategies. We differentiate strategies between those which tell you what to do and those which help you get the job done. Yellowstone hosts these very complex trading strategies, tests them and helps the trader to optimize them. Mostly what we saw was people working in Excel and MatLab.
Traders: What do you mean by "arbitrarily complex?"
Smith: We want to enable them to handle as complex a strategy as they can come up with. We can even integrate Matlab into a strategy. If there is some routine in MatLab that is so complex, that you want to call that routine,' we can do that. A lot of the other products we saw were unable to get very complex. How complex is your decision to invest? If you are doing an order handling or execution strategy, how complex is your decision how to trade?
Traders: Why was Yellowstone needed?
Smith: With MatLab and Excel, you are redesigning the plumbing every time you [write a strategy] instead of just focusing on the business logic behind the strategies. There's traders out there with a lot of brilliant ideas. They need to be able to test and go through a lot of ideas quickly to find ones that work. So, they really can't afford to be sitting there doing the plumbing every time. We wanted to put a system together where all they are really writing is the business logic and not the plumbing.
Traders: What do you mean by plumbing?'
Smith: Underneath Excel is plumbing. That says: when I change this cell, change that cell and keep track of the formulas and allow me to audit the formulas and what effects what, etc. Plumbing that allows you to plug in add-ins like Bloomberg or custom formulas. And you don't want to write all that stuff. You just want to write your formulas in Excel spreadsheets. Yellowstone lets you do your position tracking, your order tracking, market data interfaces, calculations, currency conversions, and order formation. It asks you whether you are doing constant money or constant units or shares. It does commission calculations. How does the commission effect you? When you are constructing a trading strategy you don't want to have to re-write all this stuff every time.
Traders: Will 4S.Yellowstone work with Excel if necessary?
Smith: Yes. That's a key design feature. We wanted to be able to integrate with Excel because we knew it was ubiquitous. Certainly, you can send data out of Yellowstone into Excel and take stuff in from Excel. But you can also actually write your business logic in an Excel macro.
Traders: Your second product 4S.Klondike helps traders find profitable pairs. Is most of the demand for your technology coming from pairs traders?
Smith: We saw a lot of demand for pairs trading. But we really went into it without any notions. 4S.Yellowstone would do pairs, indices, single instruments. It didn't really matter.
Traders: 4S.Klondike uses statistical analysis to find potential trading pairs?
Smith: Some potential customers knew the pairs they wanted to trade, but they also wanted to investigate non-standard pairs. Someone would trade Air France against Lufthansa, for example. They're both airlines. But what about Air France against General Electric? Does that make any sense?
Traders: The software determines the probability that these pairs will be a good trade?
Smith: It's not really a probability tool. It is looking at correlation, ratio and regression. You can look at the correlation over three different time frames: short term, medium term and long term. For example, find me the pairs that are correlated over the long term, but over the short term their correlation is pretty variable. So when the short-term correlation goes way out of whack, a pairs trader places a bet because he thinks the long term correlation is there.
Traders: OK. Go on.
Smith: Most of our strategies are either mean reversion or momentum. When something is going back to an average or a mean or they find momentum and ride it.
Traders: So, if Lufthansa and Air France generally trade within a few euros of each other and one suddenly moves sharply, the idea is that they will eventually revert to their traditional relationship?
Smith: That's right.
Traders: How have traders traditionally sought out pairs trading opportunities?
Smith: By constructing giant spreadsheets. They've either tried to find things or have stuck with the standard known pairs and quantities. Klondike allows you to put in your own statistical filters and scoring systems. You can score pairs and select them based on statistical criteria. It won't tell you those pairs make any money. You would have to place those pairs in a trading strategy and simulate the trading strategy in something like Yellowstone to see whether it makes any money. Klondike lets you cover a lot more territory a lot faster.
Traders: Someone told me recently that all the pairs were known. That you can't make money trading them. Is that so?
Smith: The known pairs are known. Whether or not you can make money off of them is not only a function of knowing the pair. It also depends on the sophistication of your pairs trading strategy. I can identify a pair that has a mean that it will revert to. And I'm really good at identifying that it has deviated from the mean. So what do I use to pick that? Do I look at regression? Do I look at ratio? To understand that it has deviated from the mean? There are still people making money in this area although I agree that the easy money is gone.
Traders: 4S.Everglades runs the strategies that 4S.Yellowstone tests?
Smith: Right. Some of our clients and prospects told us we were helping them figure out whether an idea worked or not, but were not helping them make any money off of it. So what if I know the idea works. Now how do I trade it?
Traders: O.K.
Smith: And when you get in and trade it, you encounter a host of other issues. I need to know the status of my orders. I need to be able to go out via FIX or through an API to a destination. I need to be able to run order-handling strategies. I want a strategy to get pairs done in some way. I want to get a basket done in some way. These are strategies where the content has been pre-determined either through Yellowstone or spreadsheets or MatLab or whatever. I know what I want to trade. Now help me work this trade.
Traders: In other words, how much to feed into the market at what time, etc?
Smith: When to feed it in. Can we feed it under these conditions? How to handle multiple instruments. How to deal with a legging condition. That's when one leg gets done and the other one doesn't. How complex do I want my business logic to be around that?
Traders: In industry parlance, is 4S.Everglades a "black box?"
Smith: I guess it would be. And also Yellowstone to some extent. But we don't like the idea of a black box. We believe in the transparent box. We've tried to build a high degree of transparency into all of our products. So the user can see exactly why decisions were made every step of the way. It is very difficult to de-bug a black box. Our stuff is a platform for trading ideas. Trading opportunities. So, if you have a trading idea and you want to find out why it had a certain P&L or why it created certain orders? Why did this happen? You need a transparent box, not a black box.
Traders: With Yellowstone, where does the data come from? Do you have feeds?
Smith: We are not a market data provider. We have adaptors to popular market data sources. The users have their own deals for the market data through the common vendors out there.
Traders: Do they use vendors or go direct to exchanges?
Smith: We actually have an adaptor to one of the exchanges, but I haven't seen anyone use that.
Traders: 4S.Everglades plugs into brokers?
Smith: Exactly. We connect to about half a dozen brokers through FIX. There are others we could connect to if clients wanted to go to them.
Traders: Can you mention a few of the brokers?
Smith: Credit Suisse First Boston's AES, Barclays Capital, Lehman Brothers. Fimat, the brokerage subsidiary of Societe Generale in France.
Traders: Is 4S.Everglades competition for brokers' algorithmic trading servers that execute against benchmarks such as VWAP?
Smith: Not really. The buzz around algorithms is good for firms like ours. For those customers looking to run alpha strategies, a product like ours can be used in combination with what [the brokers] are doing. They would use Everglades with an [CSFB] AES tag, for instance.
Traders: So 4S.Everglades execution strategies are different? Not just VWAP?
Smith: You can do that in Everglades. But our strategies…Do this when market conditions are such. Get a pair done in a ratio. Get an index done in this way. Get a basket done when certain conditions happen. As opposed to: get a basket done at VWAP. It is more opportunistic. I want to wait until the market hits this level, plus or minus some percent and then put in an order at some other price. I'm not saying someone couldn't write a sophisticated VWAP with our stuff. But that is not what they are coming to us for.
Traders: O.K. Thanks Steve.
