A Two-Way Street
Futures Strategies and Automated Trading
“Automation and trading strategies feed on each other.”
Talking to high-frequency traders makes you more aware of the central role of technology. Roy Niederhoffer says his shop's trading is 95% automated—the head trader watches what's going on and adds a human touch when there's an unusual event. Some managers rely 100% on automated trading.
We wanted to understand how the ongoing development of automation is affecting investment strategies and vice versa. Tom Haldes and Elise Fleischaker of Trading Technologies give us the latest perspective on the interface between automation and strategy. Mr. Haldes knows the issues from both the technology developer and user sides—he headed global market data development at Citadel Investment Group earlier in his career.
Electronic trading has opened up a host of possibilities. Modelers can choose among different levels of automation. A black box – a fully automated program – sends thousands of orders a day to markets and implements the circuit breakers and failure detection mechanisms indicated by the trading strategy without anyone lifting a finger. A grey box is a hybrid that incorporates more user control.
Opalesque Futures Intelligence: What's next for electronic trading?
Tom Haldes: Dramatic as the growth of electronic trading has been in the past few years, it shows no sign of letting up. TT is developing next generation, best-of-breed automated trading systems for high performance execution. This is a very fast changing area.
Elise Fleischaker: When TT was founded in 1994, electronic trading was in its infancy and there were questions as to whether it would catch on. In the past several years the number of markets we connect to has more than doubled. Initially we were known for our point-and-click, manual trading interface - which is still an important component in our platform - but automated trading technologies have become increasingly important for our customers and as such, that is where much of our new development resources are focused.
OFI: What kind of improvements do traders want?
TH: Probably the biggest factor is the speed of trade execution. People create black boxes and use grey box tools for automation. That's all fine, but if your trade order hits the market half a second after someone else's order and they get the price you wanted, the automation is not working for you.
OFI: Does automation create new strategies or are the strategies pushing automation?
TH: Automation and trading strategies feed on each other. For example, consider a black box application developed for index arbitrage--a strategy to gain from price anomalies between a basket of securities that is a proxy for an index and the index itself as it trades in the futures market. A black box routes an order to buy or sell, depending on whether the price of the index is below or above the theoretical price indicated by the model. This is a fairly common strategy, so likely there'll be another trader with a similar model. When the price discrepancy shows up, both traders submit orders for the index. The order that hits first gets filled before the price moves, while the later order misses the price. So everybody wants their automation to be as fast as possible. The big trend in all the markets is reducing latency.
OFI: When you say fast, how fast do you mean?
TH: The standard used to be seconds, recently it became milliseconds and now the game is being played at the microseconds level. There are systems that can respond to changes in the market and re-submit an order within hundreds of microseconds. That's how latency-sensitive the business has become.
OFI: Which strategies does the issue of latency really matter for?
TH: If you rely on automation either in part or totally through a black box to execute your strategy, reducing latency is important for you. There are traders who might have a broker manually working on their order and in that situation the milliseconds don't count. Manual trading is a different story, but if you've embraced automated trading, speed matters, and even more so in arbitrage strategies.
OFI: How is latency reduced?
TH: Take the delay in getting connected to an exchange because it is physically distant. In this case, the latency exhibited by the trading system is largely the result of the geographic separation of the order routing logic and the matching engine. Traders want to trade as quickly as they could if they were in the same place as the exchange, even if they're in another continent! We're creating server-based platforms that remove this type of delay. With high volatility in markets, milliseconds can make the difference between a successful trade and a failed one, and if you're executing a strategy on a matching engine located somewhere across the globe, you need a proximity-based solution to eliminate the geographical latency.
OFI: Would you give an example?
TH: For instance, in calendar spread trading, where you go long one futures contract and short another, the system enters a quoting order for one leg while leaning in the other market for the hedge leg. As the hedge market moves, the quoting order needs to be re-quoted to preserve the target implied spread price. If you're in London and the exchange is in Chicago, the implied spread price may change several times during the interval of time it takes to receive your market data update and respond with an updated re-quote. Thus geographical latency can make your spread trading less effective. Our popular spread trading application, Autospreader, was initially developed for desktop computers. We're developing new technology that will move the execution logic very close to the matching engine, eliminating geographical latency. This means the trader in London trading on an exchange in Chicago can see his quoting order update in less than a millisecond after the lean market has moved, just as if he were physically located in Chicago.
OFI: How does a researcher's quantitative model work with your trading software?
TH: Traders can execute their strategies using applications like Autospreader. But very complex quantitative models and algorithms may not fit any stand-alone tool. We have special applications for writing programs and creating black boxes. These are used in conjunction with our exchange gateways to submit orders and receive prices. By combining the tools, a sophisticated quant shop can build any model they want and link it to our trading platform.
OFI: What else do traders need, besides speedy order execution?
TH: In the race for performance, huge amounts of market data are being generated. Exchanges are quoting prices thousands of times a second. How will the ever-growing volume of data be managed? Traders, exchanges and telecommunication providers are all grappling with that question. If the data gets queued in the pipeline, then your system will lag in responding to market movements. So a corollary of automated trading is how to accommodate what's come to be known as the market data tsunami.
OFI: How would a shop that does not have a big IT department manage the technology?
TH: Hence there is a growing trend for hosting services. Traders and brokers want access to dozens of exchanges but don't necessarily want to manage all the lines, circuits, gateways to the exchanges, servers, etc. A hosting provider gives you high-speed access for trading markets worldwide. That way, you focus on your core business and not on the technology.
EF: Customers that use our TTNET hosting service outsource the management of their trading network and telecommunications infrastructure. We deploy, monitor and upgrade their network, including the gateways. The trend towards outsourcing is not focused on small firms--some of the largest global banks use our service.