bill and all, article in today's ny times about financial forecasting (with a jarring statement that TA will allow people to get out before the BK!)
December 23, 1997
By BRUNO GIUSSANI
Crunching the Numbers Anew
To Predict the Market's Future
ZURICH - Is it possible to forecast the future of the financial markets by
studying their history?
Classical financial theory says it isn't. Markets, it contends, are as random as a
drunken sailor's walk: no matter where he has stumbled so far, it is impossible to
predict what his next step will be. Chances are equal that he will move left or right,
or anywhere in between.
"Market makers don't take decisions based on yesterday's data; they want the most
up to date information," said Richard Olsen, the founder of forecasting firm Olsen
& Associates. The basic assumption is that while new, relevant information may
change the price of a stock or swing a foreign exchange rate, past performances
don't.
Yet Olsen claims that by digging into vast amounts of historical data and filtering
them out with powerful computers, complex software and a set of new and
audacious theoretical tools, it is possible to make visible hidden structures and thus
predict the general future direction of the markets.
"Of course, it will never be possible to forecast everything," Olsen warned in an
interview last week, "but predictions do have a value if they let us move slightly
beyond the 50-50 probability of throwing a dice."
And indeed there is a fragile but statistically significant relationship between past
market prices and their future evolution, Olsen claims. He said that his predictions
on currency markets and interest rates are correct 60 percent to 70 percent of the
time.
Credit: Olsen & Associates Ltd.
Directional forecasts provide information on market
volatility and directional pricing. Point forecasts, which
indicate the expected value of a price at a specific time, and
confidence intervals, which reflect the expected range into
which the future price is projected to fall 50% of the time,
are available for periods of one day to eight weeks.
After studying at Oxford and Zurich and working a few years as a foreign exchange
trader for a Swiss bank, in 1986 he set up Olsen & Associates Ltd. to put his
economic models to the test.
With its headquarters in an old textile mill on Zurich's lakeshore, today the
company employs about 50 people and sells forecasting systems and trading models
as an online service to banks and other financial organizations.
Versions of the service are also available to subscribers on the company's Web site,
along with a free currency converter.
There have been times however where Olsen had to "be very cautious about
voicing my theories, in order to avoid being thrown out of the room," he said.
The financial establishment scoffed as he started, almost 12 years ago, to collect
and store so-called high-frequency market data, for example, tick-by-tick quotes
made by traders worldwide, captured from real-time data feeds such as Reuters,
Kinght-Ridder and Telerate. At that time these data weren't even being stored.
Today Richard Olsen sits on the world's largest database of foreign-exchange price
data. "We collect up to 20,000 ticks a day, as much as it used to be gathered over
50 years or so," he explained.
Indeed, conventional financial analysis used to be based on daily or weekly market
information, usually using closing prices, while ignoring the hundreds of prices
quoted during the day. "This meant that you never knew what you were looking
at," Olsen said.
"In a day's time, the structural changes in the world are negligible, and players don't
change," he added. Therefore, by examining the microstructure of the data "it is
possible to discover patterns that are independent of the events - and specific to
the market's structure."
Enter Olsen's key analytic tool: fractal theory.
Described by Benoit Mandelbrot almost 40 years
ago, the theory suggests that fractal systems show
similar patterns when observed at different scales.
"If you examine a price structure on one scale, you
can infer things for a longer period," Olsen
explained. Studying high-frequency data with a
scientific approach, he and his crew discovered,
for example, that the patterns of volatility of
foreign-exchange prices at 10-minute intervals are
consistently similar to those measured at hourly, and even at daily intervals. This
scaling seems to hold for periods up to two months.
(Volatility is a measure of the frequency with which the price of an asset changes
- boldly, a measure of risk.)
Traditional economic theory assumes that volatility is uniformly distributed in
financial markets, because these are homogeneous - traders are more or less
equal, and all have access to the same information almost instantaneously.
"This is not true, markets are heterogeneous," Olsen said, "they are made up of all
sorts of different people which trade at varying rhythms."
"When an event occurs, everyone interprets it according to his own time frame," he
added. Floor traders, corporate treasurers or central bankers all have different
responses to the same event.
"The sum of their responses is a new event, which produces new responses, and so
on," Olsen explained. A primary event (say, a ministerial crisis, or the release of a
country's trading figures) "is like a stone dropped into the lake," he said looking out
of his window at Lake Zurich. "We can't forecast the falling of the stone, but we
can predict the pattern of the secondary events, of the waves which are sure to
follow."
"The markets will still go up and down in the future, but thanks to predictive
technology everybody will have an opportunity to sell before it goes down," he
claimed.
This evolution has been made possible primarily by the availability of computing
power, which allows users to store, crunch, retrieve and analyze huge amounts of
data, and of sophisticated database software.
Olsen & Associates is a technology company, not a Swiss bank - his team comes
more from informatics and theoretical physics than from finance - and Richard
Olsen is convinced that banking is quickly morphing into engineering.
The Internet, he said, will have a "devastating impact" on financial markets, "far
more profound that we realize today."
"The Internet is not good at multimedia applications, but it's great for financial
transactions," he added. By cutting transaction costs and lowering entry barriers, it
is likely to widen the spectrum of the players and allow the creation of transaction
markets "for things we never imagined -- we will have an infinite number of
markets."
This will make the global cybermarket "far more attractive than the existing
financial markets." And predictive technology will be a consistent part of it.
"It's just a matter of time before it takes off," Olsen said.
How long?
Olsen looked out of his window at the windy lake. "Three years." |
