It had a confounding track that kept edging to the east, a relatively weak punch for a storm its size, and it seemed one jump ahead of the high-tech computer models that predicted its next move.

Hurricane Floyd was an overweight puzzle of a storm, faking out thousands up and down the East Coast before it quickly passed through the eastern reaches of Virginia and Maryland yesterday.

Enormous in breadth as it approached--it brought a worried President Clinton home from Asia--Floyd toyed with the southeast coast, and, once ashore, did far less damage than earlier, nastier storms a fraction its size.

Who could have predicted?

The weather models, perhaps.

Models, which have been around for decades, have become an increasingly valuable tool in the art of hurricane forecasting, experts say, as computers have become ever more powerful.

Now, according to Stephen Lord, acting director of the National Weather Service's Environmental Modeling Center in Camp Springs, "the primary guidance for hurricane forecasts are a set of model runs."

A model is an immensely complex computer program that uses the laws of physics along with current weather data to predict the progress of a storm. Several are used by government and private forecasters.

The primary model the weather service uses is called the GFDL, for the Weather Service's Geophysical Fluid Dynamics Laboratory in Princeton, N.J., which painstakingly designed it over two decades.

For major storms, it is run four times a day on a Weather Service supercomputer in Suitland.

The Weather Service also consults a Navy weather model called NOGAPS, for the Navy Operational Global Atmospheric Prediction System, and one called UKMET, which comes from the United Kingdom meteorological services.

But GFDL is "the best model right now," Lord said yesterday.

"I think it did pretty well," he said of the model's Floyd predictions. "I think it was doing well. The GFDL model is an excellent model statistically, overall.

"It doesn't get every forecast best," he said. "But it gets many forecasts best, and it's a very consistent model."

But at least one other forecaster found that another model did better.

Elliot Abrams, a senior vice president with AccuWeather, the State College, Pa., private forecasting company, argued that a well-known European model served him better.

AccuWeather, which Abrams contended predicted a more eastward, and accurate track, relied more on a model called ECMWF, for the European Center for Medium Range Weather Forecast service that produced it.

Abrams said his company predicted a general track that was 50 to 75 miles farther to the east than the official government track. It was a more accurate prediction "and we can document that," he said.

The reason, in part, was that the European model was "hot."

Results often vary among the various models in use, and the forecasts often result from a consensus of models.

Abrams said AccuWeather, which he said doesn't have to be quite as conservative as the government, had been relying more closely on a ECMWF because it had done well predicting the erratic path of Hurricane Dennis several weeks back.

"When a model seems to have a hot hand, you pay a little bit more attention to it," he said. The European model "also did very well on the blizzard of January 1996 and the big blizzard of March 1993."

"It seems to do well on these cataclysmic-type storms," he said.

Abrams also argued that Floyd was always weaker than its size made it appear.

Its winds, for example, were described in official bulletins as topping 120, 130, or 140 mph, he said. "But when you read the [airplane] reconnaissance reports, they always seemed to be a little less than that." He also said the storm's core of really dangerous winds was not as large as expected.

Lord added that one factor in the quick weakening of Floyd, may have been that it followed some of the same ocean path as Dennis.

Hurricanes tend to churn the ocean as they travel, pulling cooler water from the depths up to the surface, he said. The water was still cool from Dennis, and that may have helped diminish the much bigger Floyd.

Modeling and science aside, though, the final call falls to a human.

"The hurricane forecaster's job is to take the [scientific] guidance and make a forecast," Lord said. "They take the guidance provided by the models very, very seriously. But their judgment is critical in making the forecast."