Alf Hutter, 27, has been to war nearly 100 times since August, running computer simulations of what a military campaign to liberate Kuwait might look like. Each time, U.S. and allied forces prevail in air and ground battles that take just minutes for the computer to complete.

Yet as news came in last week of the stunningly successful U.S. and allied air strikes in the early hours of Operation Desert Storm, Hutter realized that none of the scenarios had assumed such a dramatic lack of Iraqi response. The computer model he was using anticipated 129 U.S. and allied fighters out of action after 3 1/2 days of combat; only a handful have yet gone down.

Persian Gulf war simulations have been underway for months in desktop computers at places such as the Brookings Institution, where Hutter works, and in the vastly more elaborate supercomputers of the Defense Department and its contractors. Much news media attention has focused on projections of casualties.

"Everyone wants to have the number," said one retired general. Everyone wants to be able to say "he's right or he's wrong, or this is the way it will go, or this is the way it won't go, or better yet, the senator or the higher-ranking official is wrong because so-and-so says that the number is this and such."

One scenario, which includes an overland drive to Baghdad, projects as many as 10,000 Americans dead and 35,000 wounded. Others place the figures much lower, although one Pentagon projection went as high as 30,000 Americans killed in the first 20 days of war, according to a news report, a figure so large that a military historian said it was "full of inaccuracies and misunderstanding of the realities of modern armed conflict."

That historian, retired Army Col. Trevor Dupuy, 74, has his own simulation that projects between 300 and 3,000 Americans killed, depending on the chosen scenario. He also foresaw greater air losses but adds: "I don't change my predictions in the slightest."

Last week, he published his 90th book, entitled "If War Comes, How to Defeat Saddam Hussein."

Joshua Epstein, a senior fellow at Brookings who designed the computer model used by his assistant, Hutter, projects between 1,049 and 4,136 U.S. fatalities after 15 to 21 days of intense combat, depending on the scenario.

"None of the runs I've ever heard of make Iraq win," Epstein said. "We always win the war. That's not really the issue. The issue is, what's the price of victory? It could be very high."

Israel is another intangible: Few modelers outside the government have been able to program Israel into the equation because of the unpredictability involved.

Behind the numbers and projections, however, lies an art and science that has been shaping and changing the way military commanders make decisions and decide on strategies. Experts say it is clear that the Defense Department's classified computer war-gaming simulations have played a critical role in the preparations for Operation Desert Storm.

"Today, they had over 1,000 sorties in about 12 hours, with one single plan, with one single air tasking order. That's extraordinary," Gen. John R. Galvin, supreme allied commander in Europe, said Thursday by telephone from a military base near Stuttgart, Germany. "And to think that it was pulled off with the loss of two aircraft against a formidable enemy. It says that senior leadership, and leadership all the way down the chain, is getting good training, and it was done a lot better because of the simulations that we've run." By yesterday 10 coalition planes had been lost, six of them U.S.

Even Gen. H. Norman Schwarzkopf, commander of allied forces in the gulf, appears to be a believer. It has been reported that Schwarzkopf, by coincidence, ran a computer-assisted war game called "Operation Internal Look" simulating an Iraqi invasion of Kuwait just weeks before the actual invasion.

Since then, Schwarzkopf has programmed Iraq simulations "almost daily," according to an interview he gave last October. Reflecting the view that computers ought to assist, not control, decision-making, he added: "I'm not one of these guys who looks at the computer outcome and says, 'Ta-da, that's what's going to happen.' I know better than that."

Good simulations make good operations, particularly at twice the speed of sound. Friday, Air Force Lt. Gen. Charles A. Horner, who has been developing the plans for the Desert Storm air campaign since August, was asked at a news conference about the seemingly impossible task of coordinating the bombing missions of seven different nations.

"We have a lot of computers, and you can bring together the tens of thousands of minute details -- radio frequencies, altitudes, tanker rendezvous, bomb configurations, who supports who, who's flying escort," Horner said. "There's just thousands and thousands of details. And we work them together as one group. . . . It provides a sheet of music that everybody sings the same song off."

War is such a complex undertaking that it is only natural the computer would find a home, experts say, just as it has in the fields of economics and politics. And its value lies most in helping to analyze the true nature of combat force and its consequences.

"We've now gone to the era of computer-controlled chaos," said a retired Army intelligence operations officer as he watched the first reports of the war on television. "Everything is computer-modeled. No longer does the commander take from his own psyche the concept of war. He fashions within his psyche the empirical data that computers are spewing to him almost at the speed of light."

"It's like the choreography of a fancy ballet, where you've got a lot of principal dancers. You've got to see how many of them are going to trip over each other -- and when," said Martin Shubik, a Yale professor of mathematical economics who has written a book on war-gaming.

"This war is not being fought to validate our war games," said retired Army Lt. Gen. John H. Cushman, who participated in the early development of computer combat modeling. "But if this war is fought very successfully compared to past wars, possibly the warfare simulation program of the . . . services would have made a contribution."

For Hutter, 27, an Army intelligence reservist who may soon be called up, the computer has merely given him more insight into potential things that could go wrong. "It is kind of scary," he said.

Staff researcher Lucy Shackelford contributed to this report.