To receive the weekend forecasts of stock-market advisory services, a subscriber usually has to wait for Monday morning's mail, which is unreliable, or phone up a proprietary "hot line," which is expensive. Now, though -- thanks to the latest technology and the ingenuity of analyst William A. Finnegan of William Finnegan Associates Inc. in Malibu, Calif. -- anyone can be his or her own prognosticator by Saturday at noon. The secret: a clever device that performs its forecasting feats in the privacy of your own home.

Finnegan is an engineer, MBA, investment counselor, and Hewlett-Packard dealer, and evidently drew on all these roles to create his system. For $460 he will sell you a Stock Market Computer, which is an HP-41CV programmable calculator -- a powerful little machine in its own right -- that is preset to ingest a week's worth of market data and spew out its assessment of how things will be looking on Wall Street some 80 days later. Depending on what the numbers say, the owner of the machine is thereby advised to buy stocks, sell them, or go short.

Besides the calculator, al you need to work the system is a set of statistics that can be found in Barron's, the financial weekly. Pumping these numbers into the machine for a few minutes each week is sort of like making biscuits with a packaged mix: The cook goes through enough motions to feel that he or she has actually baked them. In point of fact, though, it is Finnegan's own private formula -- safe inside the bowels of the customized calculator -- that does all the work. The computer assigns each statistic the weight that Finnegan, who has spent 22 years reducing 180 predictor variables to the 20 "best," is convinced that it merits.

The idea follows the old and somewhat self-evident principle that a market move, once underway, tends to continue in the same direction. Nor are Finnegan's data sources original: Most are derived from old chestnuts like the Dow Jones Industrial Average and the New York Stock Exchange odd-lot activity. Still, the proof of Finnegan's system is in the pudding. The calculator's figures -- had they been available then -- would have had you acting well ahead of the herd at such important junctures as the first quarter of 1980, the third quarter of 1981, and just prior to the bull-market explosion of August 1982. Its most urgent call to buy would have occurred in January 1975, well in time for the investor to get in on the strongest bull move in modern history.

In unusual times, the program can be volatile Like all followers of price/earnings multiples, for example, Finnegan was thrown a curve by the events of the past 18 months. In late 1982, the Dow's P/E more than doubled overnight, rising to 28.8 as a result of International Harvester's billion-dollar loss. In the middle of 1983, with other earnings problems surfacing, the multiple rose to over 100, a level that ordinarily would have helped signal a bear market. But Finnegan was able to make an adjustment to the multiple and now dutifully publishes the figure each week in his Barron's ad. As other variables lose their statistical usefulness, Finnegan promises, he will devise a new formula and banish the old one every two years.

What he can't promise, of course, is that the user will pick the right stocks to buy and sell. A program that anticipates the traditional investor emotions of greed and fear well enough to help out on that score has yet to be devised -- and is bound to be considerably more costly.