DragonHorse: Trend-following in the Far East.

Kingo Tanaka, developer of the long-term trend-following program DragonHorse, which earned 102.04% in 2000, and chief trading advisor of Tokyo-based Kanetsu Asset Management (KAM), followed an unlikely road to his trading career. As a junior high school student in Japan, Tanaka closely followed

After Tanaka graduated from Tokai University in Shizuoka, Japan, he raised the money to study abroad. The locale? Tennessee University at Martin, about three hours drive from Memphis, where he studied English and accounting. But as his savings dwindled, he had to return to Japan and found a position with futures broker Kanetsu Shoji.

After helping Kanetsu set up its Singapore branch, Tanaka was assigned to the company's international department. He sat on an order desk and dealt with overseas customers from South Asian countries such as Singapore, Thailand and Malaysia. While there, Tanaka noticed similar tendencies among losing traders. He took that observation a step further, statistically analyzing their characteristics. He discovered two distinguishing traits: they traded against market trends and used poor risk management.

Tanaka first used this knowledge to develop trading systems for his customers.

"I did not want to see my customers lose money and wiped out from the market any more, and I thought I might have helped them with my positioning advice derived from my system," he says, adding they unfortunately rarely took that advice.

But that hurdle inspired Tanaka to pursue another opportunity. In 1994 he persuaded his boss to let him trade a proprietary account for Kanetsu using his system. That first generation, however, provided only mediocre performance, earning a 10% annualized return at most using moving averages. Tanaka improved it by adding other technical methodologies.

"I stuck to trend-following," he says. "It was a common view that trend-following does not work in Japanese commodities market; however, that was a conjecture never proven scientifically. If so, I thought I would challenge the market with trend-following."

In the meantime, KAM was established in 1994 and Tanaka moved to that firm as a trader. His first system, called ELIS for Equity Leveraged Intelligent System, began trading in February 1995. In August 1998, that program evolved into DragonHorse, which is named after the Chinese imaginary animal with a dragon head and horse body and used to represent an outstanding person. DragonHorse differs from ELIS in terms of profit tendencies, position holding time and markets traded.

The program continues to be based on the trend-following concept, though Tanaka no longer uses moving averages. DragonHorse is composed of two technical techniques: an oscillator-based indicator designed to detect certain price patterns and then the recognition of formations in price itself.

Tanaka says that DragonHorse is unique, though, in the markets it trades.

"DragonHorse trades solely 19 commodity futures markets [available in Japan] such as grains, softs, metals, energy products and others, including U.S. soybeans, corn, red beans, coffee, raw sugar, rubber, silk, dried cocoon, etc.," he says. "Because of this, the system's results are generally uncorrelated with non-Japanese commodity trading advisors."

Another important feature of DragonHorse is its ability to vary the position size it takes per trade based on different levels of volatility. Each commodity is allocated a fixed initial percentage of the total portfolio without risking more, than 2.5% of an account's equity. Exceptions are allowed, though.

"Trading decision is made 90% of system and 10% of discretion," Tanaka says. "Position sizes and buy/sell signals are a system part, but the selection of markets is within discretion. And in terms of risking amount per trade, it may be expanded about three times of initial limit only when I am convinced a trend is there. Also, I use stop orders, but the levels I put stops at vary depending on the circumstances with a consideration of three elements: equity, markets traded and price fluctuations."

DragonHorse earned 46.18% in 1998, 5.49% in 1999 and 102.04% in 2000. It gained 11.12% in January 2001. The biggest monthly drawdown was -15.09% in September 2000.

Tanaka says another benefit of the trend-following nature of the program allows it to capture large profits on individual trades.

"I took more than 5,000 ticks with arabica coffee in June of 1999 and another 4,800 with raw sugar.... They are the trades I cannot forget," he says.

Euchi Kojima is a reporter for MK News in Tokyo and editor of the Japanese version of Futures.