Futures Traders Take On Next Hot Model By ADAM BRADBERY February 19, 2007 LONDON -- The biggest trader in one of Europe 's main futures contracts isn't a person -- it is a machine. So is the second. And the third. Trading systems that use complex mathematical codes to quickly weigh a huge number of possible trades and execute them faster than a human are squeezing the independent trader as they change the way futures contracts are traded. These big-ticket trading strategies are based on algorithmic trading systems, which have become hot properties for hedge funds and banks' proprietary trading desks during the past few years. They are lifting exchanges' trading volume and revenue, and reducing swings in prices. They are causing some locals, as independent traders are called, to abandon dealing in certain contracts because they can't compete. Locals are trying to fight back by using their own automated trading systems. These traders, who tend to trade in smaller amounts, are important to futures markets because they are more willing to take the opposite side of trades that reflect investors' long-term view, figuring they can profit from short-term swings. That helps keep trading liquid. On Euronext.Liffe, the London derivatives exchange owned by pan-European stock-exchange operator Euronext NV , some of the 2,000 locals contend that these systems are reducing the exchange's overall efficiency because they can exploit exchange rules that reward bigger trades. Euronext.Liffe, the second-biggest futures exchange in Europe by volume of contracts traded, says the top three traders in its flagship euro interbank offered rate, or Euribor, three-month interest-rate futures contract are machines, and algorithmic trading accounts for 20% of turnover in the contract. It doesn't have comparable figures for a few years ago, but a spokesman says there has been "tremendous growth" in this area. German-Swiss exchange Eurex AG, owned by Deutsche Börse AG and SWX AG, estimates algorithmic trading makes up 15% to 20% of its overall business. U.S. futures exchanges say they are experiencing big growth in model trading. The Chicago Mercantile Exchange says activity from models is growing on its exchange but it can't track the amount. Some locals trading on Euronext.Liffe say algorithmic models are exploiting the way Liffe's computers match buy-and-sell orders for short-term interest-rate futures, which gives a larger share of incoming business to the biggest orders in the market. This largely pro rata system avoids having orders held up at the back of a queue because they were placed later and is designed to spread business across a range of orders in the queue. Eurex uses the same methodology for its Euribor futures, but Liffe has the bulk of business in this contract. Both exchanges use a price-time methodology for longer-term bond-and-stock contracts, which focus largely on a first-come, first-served system. A Liffe executive said the exchange is reviewing its rules in response to complaints from some market participants. Here is an example of how Liffe's order-matching system can work: An algorithmic trading model places an order to buy 20,000 contracts of a short-term interest-rate contract at a set price on Euronext.Liffe. That order represents the bulk of the orders at that price. A local also wants to buy 100 contracts at that price but neither of these orders was the first to be entered in the market at that price. When a third trader offers to sell 100 contracts at that price, the exchange usually automatically allocates most of that order against the 20,000-contract order, even if the local's order was placed before the larger trade. The big buyer's trading program then quickly pulls the balance of the order. The trading strategy doesn't violate Liffe's rules, or those of Deutsche Börse, because the algorithmic systems could, theoretically, be hit on their entire order. Because Liffe's own matching system is so complex, smaller traders aren't always disadvantaged because of the size of their orders. The strategy makes it difficult for locals to get their trades completed. It also can leave them with partially filled orders, which in turn makes it difficult for them to trade on the relationship between contracts with different expiry dates, known as spread trades, as many locals do. Many computer models are competing with locals in a strategy known as scalping, which involves making small profits from the gap between the bid-and-offer prices. The speed at which these systems operate, and their sizable orders, means that locals are finding it harder to compete. Many observers say locals need to develop new trading strategies and adopt fast automated trading tools to compete more effectively. "The growth of algorithmic trading is inevitable and locals need to adapt to it and develop programs to complement their own trading," said Peter Green, chief executive of Kyte Group Ltd., a trading arcade that provides trading systems and services to locals. Some locals say the systems' activity of entering large orders and then pulling the bulk of each one also creates a false impression of how much interest there is to buy or sell contracts, which is known as liquidity. "The quality of the Liffe market has been depreciating because of the growth of the machines," said Todd Van Hees, a local at Futures Spread Trading, a company that provides services for independent traders. A Euronext.Liffe executive said the exchange is reassessing the impact of its pro rata matching algorithm to see whether it treats all customers fairly. "We know there are some concerns about the algorithm," said Amanda Sudworth, Euronext.Liffe's director of interest-rate product management. "We are exploring enhancements to our algorithms to make trading more attractive to all traders, taking the changing dynamics of the market into account. We will be making announcements when we have finished our consultations." Some locals have complained to the exchange about the activity of the algorithmic trading models. According to an active trader of the Euroswiss future, one of the exchange's smaller interest-rate contracts, the exchange management recognized that a new algorithmic trading system was having an impact on price action in that contract and that this was discouraging locals from trading. Ms. Sudworth said there hasn't been a particular problem with that contract. Those locals who remain on Euronext.Liffe say they are having to enter bigger orders in order to compete with the trading systems. That means they are taking on more risk and can end up trading more contracts than they want to. They say this is making it harder for new locals to start trading in the market because they don't have the leverage to compete with the size of the systems' orders. The exchange says locals make up 45% of its Euribor futures trading and this has remained broadly stable during the past couple of years. The exchange declined to comment on the figures for Euroswiss futures trading.
"They are lifting exchanges' trading volume and revenue, and reducing swings in prices." 20 % of volume done by machines, makes the movements less "human" , not a good thing, I prefer human fear and greed driven movements > better to recognize and take advantage off.
The exchange wants to see volume grow. If volume leaves because locals in effect "strike" then pressure will grow to change something. The algo they use for the 5 year notes at the cbot has driven locals out of those markets. The absolute last thing anyone would want is markets dominated by a few massive programs because when the inevitable happens it will be massive. Too much risk on too few participants.
Link or reference? Curious to find out more about that.. The 10-year (ZN) does not use the same algo?
http://www.ecbotdirect.com/cbot/pub/cont_detail/0,3206,115+42577,00.html The 10yr is still on first come first serve, queue system, 2 & 5s have gone to pro rata. See the pdf in the link above for more details Actually having skimmed through that article I linked, it looks like they've now reverted back to FIFO in the 5s
The article mentions a reduce in swings, did the same thing happen also in other instruments ? To me it seems a logical consequence of computerized trading. What you think ?
I agree, I think when the markets were first tackled on screen by computer it caused the big swings with automated stops being triggered. Big stops are much rarer now, the markets used to shut down in eurex due to too big a move, on something like non farms. Now that there are so many computers in the markets place and rapid automated scalping software, any big out of line moves never really get anywhere before being mopped up instantly by the fastest program to see a disparity. The size of the bids and offers has increased massively in something like the tnote. You're looking at 6-10,000 a side now whereas 5 years ago it was just a few hundred. I do wonder how big these orders will grow before something hits a limit, with so much money washing around the markets at present I can't see it slowing any time soon. If a black box system is profitable I guess you order it to double its trading size at every profit target accomplished, I'm not sure what will eventually snap?