neural networks

From: maness@sdcoe.k12.ca.us (Jay Maness)
Subject: neural networks
Date: Fri, 7 Aug 1998 05:02:56 -0700 (PDT)
Next Article (by Subject): Re: persfin-digest V5 #41 Mortgae for Mutual Funds GeorgeS
Previous Article (by Subject): looking for Expenditure verses Income Level statistics Tim Glaze
Articles sorted by: [Date] [Author] [Subject]
<Memills@aol.com>  asked:

Subject: Artificial Intelligence/Neural Net based mutual funds

Does any know of a mutual fund that makes it stock selections
based solely on a neural network (artificial intelligence) analysis
of a historical database?   I'm familiar with James O'Shaughnessy's
work, but I don't think he used neural network software.

Thanks,

 -- Mike Mills


<reply>

Mike,  I have some interest in this field also.

Here is a lightly edited copy of a posting from MONEY that I kept.

<begin quote>

Date:         Fri, 8 May 1998 20:13:27 EST
Reply-To: dailymail@LISTSERV.PATHFINDER.COM
Sender: Money Daily <MONEYDAILY@LISTSERV.PATHFINDER.COM>
From: MONEY Daily <moneyadm@PATHFINDER.COM>
Subject:      How one fund capitalizes on market inefficiency
To: MONEYDAILY@LISTSERV.PATHFINDER.COM

For an enhanced HTML version of the Money Daily,
visit http://moneydaily.com.

Weekend, May 9-10, 1998

How one fund capitalizes on market inefficiency

The new N/I Larger Cap Value Fund trades hard and fast
on a fair value model that's recalculated hundreds of
times a day - and so far, it's working

By Michael Brush

In business school, they teach you that the stock
markets are completely efficient. All public
information about a company, the professors tell you,
is fully reflected in a stock's price at all times.

Rubbish, say many seasoned investors. They know that
stock prices often bounce around maddeningly, even
when there is no news. And when there is news, look
out! Investors are prone to overreact, driving prices
to extremes.

Capitalizing on these temporary price anomalies would
seem like a winning strategy. A relatively new mutual
fund that tries to do just that is the Larger Cap
Value Fund (NILVX), launched by N/I Family of Funds (800-
NUMERIC) last December. The fund is based on the
premise that when it comes to the stock markets,
business school professors have it all wrong.

"The markets are not efficient. They are partially
efficient," says Langdon Wheeler, who founded N/I in
1989, and helped develop the computer programs behind
the Larger Cap fund. "It is not all machines yet. The
markets are people and people get carried away. They
are out there saying `I have to buy because he is
buying.' Or `I have to sell because he is selling.'"

And that kind of emotion often helps drive down prices
of good stocks to attractive levels. "Prices slosh
around quite a bit," notes Wheeler, an former engineer
who became fascinated by financial markets and changed
careers. "Just look at a price chart. You see all
kinds of ups and downs when the fundamentals did not
change that much."

To take advantage of buying opportunities caused by
market inefficiencies, the Cambridge, Mass.-based N/I
has developed a `fair value model' which tries to
"mathematically model the excesses of human behavior
in the market," says Wheeler.

The first step is to figure out the `fair value' -- or
how much the market is usually willing to pay -- for
certain qualities of stocks. These include things like
book value, expected earnings, the growth rate, the
quality or consistency of earnings, and the number of
analysts covering a stock. The market, for example,
tends to pay less for volatile earnings as opposed to
more predictable earnings. And it pays more for
companies that are followed by more analysts.

"You figure out how the market prices all these
factors each day," says Wheeler. "Then you go back and
assign each factor a weight for each stock, and bingo!
You have a theoretical price." The computers at N/I x-
ray the stock market about once every 90 seconds to
see if investors are underpricing any aspects of a
company, like its earnings growth or book value. If
so, portfolio manager Arup Datta and N/I traders
pounce on the stock.

As you might suspect, such a portfolio management
style leads to heavy trading. The fund has about a
250% annual turnover -- way above what many fund
advisors believe is an acceptable level. Two factors,
however, make this high turnover rate palatable.

First, the fair market value strategy creates a good
enough return to offset the cost of all that trading.
As of March 31, the Larger Cap Value Fund had a return
of 16% since it was started at the beginning of
December, compared to a Russell 1000 return of 11.6%.
And over a full year, use of a similar model in
private accounts has produced the same kind of edge.

Second, N/I has a record of keeping down other costs.
Like other mutual funds at the group, the Larger Cap
Value fund has annual expenses capped at 1% -- below
average for an actively managed fund. And there is no
load or 12(b)-1 fee.

Sometimes, of course, the model leads fund managers in
the wrong direction. Stocks, after all, do sometimes
go down in value because of bad news that is not yet
out. "You know you are going to be wrong on some of
them," admits Wheeler. "But you have to follow the
discipline, if you have a discipline that works most
of the time. A relatively small informational
advantage applied over a large number of trials will
get you a persistent outcome."

One of the advantages of the system is that it leads
the fund to low valuations "in stocks that most value
investors would not understand," says Wheeler. By that
he means that it finds bargains not only among the
typical down and dirty value stocks, but among quality
growth stocks as well.

If you are interested in the fund, don't wait too long
to look it over. To date, it only has about $15
million in assets. But N/I, which has about $5 billion
in assets under management, has a habit of shutting
down funds to keep them manageable -- and this one is
scheduled to close at $500 million.

<end quote>





- -
Go to Jeff Salisbury LWGate Home Page.