Hollywood doesn't announce box office results until midday each Sunday. But by dinner time the Friday before, any show business executive worth his car allowance can predict with 90 percent accuracy what movie will be No. 1, and how much it will make.

Over the last two weeks, for example, prognosticators have been saying that Chris Klein's "Rollerball" would somehow open ahead of Arnold Schwarzenegger's "Collateral Damage" this weekend. None of them have even seen the films. With that kind of foresight, shouldn't these people be out at the races picking ponies?

If the seers had the same kind of research at the track that they enjoy at the studios, they just might. Moviemaking is governed by equal measures of audacity and caution, the former applied when a movie is being started and the latter invoked as soon as it is finished. Nobody knows if an untested director like Peter Jackson can make "The Lord of the Rings," but the odds are weighed, the dice thrown, a prayer offered. The prudence kicks in when the finished movie is tested, advertised and released. Coming-attractions trailers, TV commercials and even movie posters are relentlessly analyzed in front of consumers, then recalibrated and tested again. The same is true of the film itself, which may be recut, reshot and rearranged a dozen different ways until preview audiences give it as high a score as possible. As soon as the first advertisements come out, the whole process starts anew, as hundreds of moviegoers are polled about what films they have heard about and want to watch. But unlike all the previous stages, this mathematical survey foretells both fortune and failure.

The audience studies are called "tracking" reports, and they have become the crystal ball for movie marketing. Three times a week, a Los Angeles organization called the National Research Group (some studios use L.A.'s rival research firm, MarketCast) telephones about 400 random Americans, asking them not only what movies they are aware of, but also which they intend to pay to see. The data is sorted demographically into six different age brackets (the youngest is 12-16 and the oldest 45-59, as the senior set apparently doesn't count), with additional categories for gender and race. Every new movie in wide release pops up on the surveys three weeks before its debut. In the final days before the movie opens, NRG estimates its first weekend's box office. Each studio spends hundreds of thousands of dollars a year paying for the surveys, scrutinizing the votes the way the Gore team did Florida ballots.

The numbers are invaluable because they reveal who is interested in a movie and, more important, who isn't. That data then determines how advertising money should be spent, and whether new commercials should be deployed. The tracking for "Collateral Damage," for example, reveals less popularity among young males than "Rollerball," and that's the reason many expect the latter to perform better this weekend. Specifically, according to the NRG numbers, of the people who say they will definitely see a movie this weekend, 30 percent of males under 25 favored "Rollerball," while only 19 percent picked "Collateral Damage." The numbers were even more lopsided among young women, who prefer "Rollerball" 14 percent to 4 percent.

Likewise, Denzel Washington's "John Q." is tracking much higher among African-Americans than Bruce Willis's "Hart's War," suggesting that Washington's movie may debut stronger than Willis's film on Feb. 15. (It also shows why MGM and New Line discussed New Line's postponing "John Q." to settle MGM's fight with New Line over the title "Austin Powers in Goldmember," a play on MGM's James Bond movie "Goldfinger.") Young men have scant interest in "Dragonfly," only confirming that star Kevin Costner's audience is aging even faster than the actor. To goose a movie with scant teen interest, a studio may purchase more commercials on MTV. If women haven't heard of a title, the advertising buy might shift to "Oprah." To lure a young audience to "Hart's War," where tracking surveys show the movie needs help, MGM's ads for the film make it look like a pure Bruce Willis action thriller, not the Colin Farrell prisoner of war drama it really is.

Of course, even the best interpretation of tracking numbers can't save a movie no one wants to see, which is why for some marketing executives the surveys have grown into both blessing and curse. As one executive explains, some movies are just not meant to be made. But they are filmed anyway, and-surprise--the tracking surveys indicate not a soul cares. But the film's producer, confronted by the listless survey results, insists that the studio need only spend millions more advertising the turkey to improve the numbers. Good marketing money is thus thrown after bad, and you still have "The Majestic."

The barrage of Super Bowl movie ads dramatizes the importance of tracking surveys. Even though tracking reports aren't really relevant until three weeks before a movie opens, studios are jockeying now to guarantee than when their summer film debuts on the survey, it's the first choice among fans. Commercials before, during and after Sunday's championship touted more than a dozen movies, some plugging films that don't open for six months. The summer releases advertised on Super Bowl Sunday included "Signs," "Men in Black 2," "XXX," "Spider-Man" and "Austin Powers 3." The most surprising of this group is "XXX," Sony's action movie from the director (Rob Cohen) and star (Vin Diesel) of "The Fast and the Furious." The other films showcase event--status pedigrees--"Signs" stars Mel Gibson, and was directed by M. Night Shyamalan, the maker of "The Sixth Sense." To jack up "XXX's" profile to the level of "Spider-Man," Sony believed it had to spend $1.8 million on a 30-second Super Bowl spot. When late-summer's tracking reports are released, we'll see if it was a good investment.

Like "A Beautiful Mind," the tracking surveys don't always get every detail right. NRG predicted "Traffic" would open to about $8 million; the film did almost double that. The predictions for the debut of the Bruce Willis film "Bandits" were faulty as well, as it did worse than projected. All the same, a combination of NRG estimates and some early box-office returns can prove eerily infallible. Last year I attended a Friday dinner party with a marketing chief whose studio was releasing a new movie that weekend. Like someone with a badly disguised drug problem, he got up from the table every five minutes to check on ticket-sales data from a handful of East Coast theaters. Before dessert was served at 9 p.m., he announced the movie would gross between $15 million and $16 million and land in first place. On Sunday afternoon, when the numbers were released, the final tally: $15.8 million, good enough for first place.