The craziest signal of change in the exchangetraded market ever has to be the excitement generated by the discovery that physical location closest to the wall of the NYSE shaves off enough processing time to make a significant difference to profitability. It seems somehow ironic that developments in electronic trading initially intended to enable trading from a distance have caused a bun fight for position closest to the exchange.
"I can't say I'd ever get attracted to that kind of trading, but I might at some point get interested if we can focus the necessary resources to successfully 'fight for turf' in that domain," says Bob Pardo, founder of CTA Pardo Capital Limited and system builder, mentor and coaching organisation the Pardo Group. So far, he doesn't have to fight for turf at all: his trading program XT99 has posted a gain of 225% since inception in June 1999.
Go with the flow
But Pardo thinks technology will transform the industry to the extent that many in the old school will not be able to continue to compete as effectively, which will negatively affect their future performance. Pardo works with a multi-market and quasi multistrategy system at the moment and will expand the multi-strategy element and introduce multiple time frames in the near future as he develops a hedge fund for launch in the next 18 months. He also has plans to develop an intelligent expert trading system based on a former project.
The increase in automation will vastly increase what can be extracted from the market, says Pardo. If the amount of profit available from a daily chart amounts to multiple hundreds of percent per year, say 500% per market per year across 50 markets, even the top-performing CTAs look inefficient by comparison. "I have been in the top ten many times for making a 35%+ three year compound return… and this is nothing in comparison with what is possible," he says. "The potential is gargantuan."
However, despite plans to build an intelligent expert trading system, Pardo muses on the fact that software has so far not kept up with developments in hardware: "We have a 17,000% increase in processing speed since 1990, but testing software has not increased by even a factor of ten," he says. "There is no reason this power could not be harnessed, because the technology already exists, and once it is you will see much higher levels of return with lower risk from those CTAs and hedge funds who exploit the current technologies more fully."
John Holsinger: "Humans have not changed in the last 4,000 years and neither have the markets."
RTS Realtime Systems CEO Steffen Gemuenden says CTAs are already becoming more sophisticated in what they are doing. As the industry institutionalises, they need to increase return and reduce risk both as a stand alone investment vehicle and to prove themselves in the context of a fund of funds or commodity pool: "It is probably safe to say that if a CTA is automated they are using a model of some sort to determine which contracts they will participate in and when. And, electronic market development will only increase the possibilities," he says.
StrategyRunner CEO Anna Becker agrees. "Automation has already changed the balance among CTA strategies," she says. "There are more daytrading types of strategies used by CTAs over the last three years, and cross exchange trading is increasingly undertaken by mid to large CTAs."
So many firms are now involved in trades based on price discrepancies that the trading game is changing. More and more people are looking to profit from these very small fluctuations and mispricings.
"You can sit there all day and make a penny every time there's a price discrepancy because exchange A for some reason doesn't react as quickly as exchange B, making your money out of huge volumes," says John Holsinger of money manager and system subscription service TraderSpeak.
How much and why?
While the exchanges do not have hard volume figures for CTA automated trading activity, they do have an idea of how much volume is generated by automated trading as a whole. Eurex estimates this figure to be around 20% of its turnover, with the proportion highest in highly liquid products like the Euro Bund futures or the DJEuroStoxx50, where it stands at about 50%. The exchange says the development boosts liquidity and further reduces spreads.
Amanda Sudworth, Euronext.Liffe's director of interest-rate product management believes the most sophisticated traders represent a 'quite high' proportion (up to 14%) of volume. (All automated trading tools have the be registered with the exchange). Euronext.Liffe has not so far seen an influx of super high-frequency trading; an area it expects to grow, rather than mushroom. According to a spokesman at Eurex, there are three basic reasons for this increase in automated trading: speed in interpreting market data and millisecond reaction time, objectivity (and no nervous breakdowns), and scalability that enables diversification by product and geography.
Thus the biggest uses of automation are diversification by market and spread trading, for example, of the closely-linked European markets. The general consensus is that manual arbing is no longer possible. "Comparing manual versus automated trading is like comparing a human with a jaguar," says Carl Persson, CEO of CTA One World Capital.
However, adds Pardo, "The biggest reason people use automated trading systems is because it's the only way you can properly measure risk and return in order to properly capitalise an account for maximum performance. The other reason is to extend the model to as many markets as possible. Once you're trading a few thousand stocks and all futures markets in multiple time frames you can't handle it otherwise."
RTS Realtime Systems' Gemuenden says, "You will see many different nuances of trading style within each strategy, such as markets or time horizon. Automation has definitely helped increase the ways a CTA can differentiate itself either through increased frequency of executed strategies or better quality in execution."
Strategy Runner's Becker agrees: whether CTAs use automation for high frequency trading or to increase diversification by market, timeframe, parameters, or models, automation makes the trading process more efficient, she says.
"For high frequency it's essential that systems will be running from a location next to exchanges or ECN's."
But in general, despite the availability of what Euronext.Liffe's Sudworth calls some 'cool bits of kit' that allow the less sophisticated IT people to build more sophisticated software programmes, take up is less rapid than the pace of development.
Partly, this is because one end of the automated continuum has an unshakeable belief in what Holsinger calls 'the basic stuff': trends and breakouts (the 20-day breakout of a high or low still works), fake breakouts, careful attention to the ratio and magnitude of winning to losing trades and stop mechanisms. The 24-hour market has not changed that: Holsinger's average holding period remains 3-6 months for longer term trades, with very short term trades at the other end and nothing in the middle, though his day-trading system has an extension to the following day.
Holsinger's trade efficiency is 60%, he says, and he finds, "The more I try to increase that, the less efficient the system becomes."
"I don't think it's diversification if you trade the same market using different time frames," adds One World Capital's Persson. "The same goes for different parameters in the same market. The system should be such that it can be traded in several markets without major modifications to fit the market… I don't think frequency should have anything to do with it: accuracy is the key."
Though some CTAs are using options pricing theory to construct their position building, automated trading is still easier in futures than in options. However, Euronext.Liffe plans to extend its ELPS options trading software developed for the Amsterdam retail market to its equities and interest rate contracts.
RTS Realtime Systems Gemuenden adds, "Liquidity will really dictate whether the opportunity not only to enter an option market is there but, more importantly, whether you can keep the profit when you want to close out the position."
Tools of the autotrade
Many CTAs find off-the-shelf solutions for building and deploying their automated trading models inadequate for their needs, which is why StrategyRunner's Becker says she has automated a wide range of popular trading tools as well as Excel and neural networks.
But it is difficult to generalise. Size and orientation affect choices. Traders that come from a programming background might build their own, whereas those that come from the financial angle are more likely to customise. And RTS Gemuenden says that not only the specific needs of each manager, but where he has reached in his lifecycle will dictate choices. "When they start looking at the available tools they realise that each system comes with different features and costs will vary accordingly, especially if they require specific customisations. But, in order to survive, they need to have positive returns on a consistent basis whether they are a newer fund or one that has a longer track record so having the best, affordable tools in their arsenal is paramount." Common shortcomings in popular off the shelf tools include lack of speed, absence of batching capabilities for automating testing and no portfolio or walk forward testing functionality. While there are more expensive, higher end products with good reputations, even they have their limitations. According to Pardo, "In the long haul you still have to build something yourself. You have to be able to blend multiple equity streams, do walk-forward analysis, portfolio analysis, and employ sophisticated things such as genetic algorithms, etc, and that is where most of the systems break down."
The growth in automated trading by CTAs and others has prompted exchanges to increase capacity in order to cope with ballooning traffic levels. For example, Eurex (in common with Euronext.Liffe) has made several technological developments last year within its 'technology roadmap' to accommodate automated trading.
It quadrupled the bandwidth for member connections from 256 kbit to 1024 kbit and has reduced the average roundtrip time for orders and quotes by two-thirds for the whole network; measured figures for latency stand at 15 milliseconds for Frankfurt-based trading systems.
Steffen Gemuenden: "...in order to survive, they need to have positive returns on a consistent basis..."
It also changed its software to allow member firms to subscribe to un-netted price information, with a 20% take-up. The next major step will be the introduction of Enhanced Broadcast Solution, intended to further improve un-netted market data and full order book depth, and allow customers to receive low-latency trading data according to their individual needs.
These developments will enable Eurex to process transaction volume of more than one billion quotes a day, well beyond the 2006 average of 120m and the 2006 peak of 230m per day, it says. Euronext.Liffe's similar project, Capacity Roadmap, is also designed to ensure that turnaround times do not deteriorate even if volume explodes.
TraderSpeak CEO John Holsinger reminds us that in 1993 it took 20 minutes to get a fill. "Seven or eight years ago it was difficult to get in and out several times a day without your broker wanting to kill you," he says. And as we noted earlier, exchanges and software providers both take colocation seriously. "Physics is something you just can't get around," says Euronext.Liffe's Sudworth.
"It's important to a CTA that the systems they employ offer the most efficient connectivity," says RTS's Gemuenden. In addition to model building features, RTS's Tango platform gives direct access to major markets via a data centre colocated to the markets it covers.
"For high-frequency it's essential that systems will be running from a location next to exchanges or ECN's," adds Strategy Runner's Becker.
Eurex also provides co-location 'Proximity' services, in conjunction with partners IXEurope and Colt, to help member firms optimise their transaction response times. The offering enables financial trading organisations to place their order management systems as close as possible to the exchange's trading infrastructure, so they can exchange data and process orders with minimal latency delay. CME's colocation facility, CME Local Network (CME Lnet) offers customers a CME-certified trading application in one of two specified colocation facilities directly connected to the CME fibre optic transport of the Globex network.
Plus ça change?
However, despite all the changes, the bulk of the CTA industry remain trend followers convinced that the more things change the more they remain the same, happy to harness the diversification and efficiencies new technology brings, but not necessarily to change the way they trade. "The people saying the markets are different now are wrong," says Holsinger. "When people tell me this I remind them of fear and greed: "Humans have not changed in the last 4,000 years and neither have the markets."