Want to know why our annual New Year's predictions have been an amazing 93 percent right for eight years running? Here's the secret.

In the computer industry, it takes about three years for research engineers to make a technological leap. It takes another two years to turn the breakthrough into viable product ideas and yet another four months to hammer and solder product concepts into saleable commodities.Once we know the breakthroughs (by following pure computer science) and find out how companies plan to use the new technology, voila! Predicting is as plain as meat and potatoes.

Well, not quite. We filter what's being planned through two sieves: what we'll buy and what you'll buy. (You tell us that in letters and face to face at our seminars.) If we know a product's coming down the pipeline that neither you nor we will go for, we're not swayed by whether the maker is IBM, Apple or an aggressive little upstart.

Let's look at our December '89 predictions. We said that computers with the i386sx chip would be the hot sellers of 1990. They sure were. We promised no new Mac until the end of 1990; Apple's just beginning to ship now. We predicted renewed buyer interest in image scanning, including optical character scanning. And we said color desktop printing would start to be affordable this year.

In earlier years, we picked some real dark horses, beginning with NEC computers and printers in 1983. The first portable we predicted success for was a Compaq, back when nobody'd ever heard of the company. We warned of gloomy sales and early deaths for Apple's Lisa, IBM's PCjr and IBM's first portable (the Convertible). That's what happened despite expensive advertising campaigns.

We cautioned in '87, '88, '89, '90, and still caution readers not to rush to IBM's PS line (PS/2 or the new PS/1). Here are some more predictions for 1991.

Look for the i386sx chip, not the i386, to be the dominant seller in the IBM compatible market. And a good thing, too. 1991 will still see few rewrites of good old programs that use the i386 chip's ability to pull in and push out 32 pieces of data at one gulp.

There will be moderate sales of i386 computers. Users of UNIX, large networks, and programs that do enormous math crunching will continue to upgrade slowly to those machines. But don't expect them to come down much in price until they really take hold in 1992.

Even though sales of i486 computers will be as modest as during 1990, watch for a splashy announcement by chipmaker Intel of a i586 chip. Small buyers, finally computer-savvy, will realize that relatively few companies can use that much power right now.

We're watching closely a legal tussle between Intel and Advanced Micro Design (AMD). Until now Intel was the sole maker of processing chips for upscale IBM compatibles. If AMD continues to win courtroom skirmishes, they'll make and market 386 and 486 chips, bringing prices down.

Microsoft's Windows will gain substantial market share during 1991, but will end the year far short of being the dominant IBM compatible environment.

Only at year's end will there finally be a rich supply of programs that put Windows to good use. Watch for more companies to join Zenith in bundling Windows with computers and for 1992 to be the year Windows becomes the major desktop software environment.

UNIX, a multitasking and multi-user operating system, will continue to replace MS DOS and other operating systems during the first part of 1991. But then the news will spread that you can run both MS DOS (and Windows with that) and UNIX (and X-Windows with that) on the same computer system. By late '91, desk-top users will start opting for the best of both worlds.

The year does not bode well for Macintosh power users. Ditto for Apple. We've heard serious rumbling about bugs in the memory of Apple's latest greatest Macs. That could sour sales until summer of '91. Adding this news to the fact that many last-generation Macs suffered chronic hard disk controller problems, we expect many once-burned corporate buyers of Apple IIIs, Apple Lisas and underpowered early Macs to sit back and wait so they don't get burned again.

We'd like to see Commodore's Amiga take off in design and engineering markets and the NeXT grab hold in shops that function like think tanks. But we don't expect to see either make much more of an inroad in general business use.

1991 will be the year there ceases to be much difference between the Mac and IBM-compatible worlds. Easy data swapping between IBM programs will come at last. Windows provides one way of doing it. Another is a programming process to watch, called hypertext on the Mac and smart documents or hot links on MS DOS.

CD ROM drives will become commonplace during '91. And its cousin, the WORM (Write Once Read Many) will also be ahustling.

An i486 computer with a jukebox of WORM disks can have more processing power and more storage capacity than many mainframe computers!

Now that most public phone lines can transmit 9,600-baud data, public suppliers of data transmission lines and on-line data bases will be opening 9,600-baud service. Most modem users will switch to 9,600-baud modems during 1991 to cut escalating phone costs. We predict a big price drop in these modems.

Under-$1,000 laser printers will also be among 1991's computer bargains. But we don't expect big sales. Due to the economic slowdown, a lot of companies will make do with dot matrix printers.