Throughout Israel's election campaign, polls have consistently shown the new Kadima Party founded by Ariel Sharon far ahead of its rivals. So why are pollsters getting nervous with the vote just days away?

For starters, survey-takers use voting patterns in previous elections to help them in their predictions, particularly about what undecided voters will do. But Kadima was created only five months ago, and its dominance has cast doubt on some of the assumptions crucial to the pollsters.

"This is a profession of assumptions," said veteran pollster Mina Zemach.

Kadima, cast as a centrist force anchored on Sharon's personal popularity, immediately leaped to the top of the polls with about 40 seats of the 120 in Israel's parliament. Labor trailed far behind with about 20 seats and Likud — which Sharon left to form his new party — limped in third with about 15, down from 40 in the last election.

Sharon suffered a stroke on Jan. 4, and the militant Hamas group swept Palestinian parliamentary elections Jan. 25. Yet neither major event caused a ripple in the polls.

Pollsters are wondering what's going on.

"All this gives me the impression of quicksand," said pollster Rafi Smith.

The main added variable is Kadima itself. It sprang from nowhere into the lead. Previous attempts to forge centrist parties in Israel have ended in flashy failures. Now the pollsters have to readjust their assumptions — or throw them out.

With predictions of a turnout of less than 60 percent and polls showing more than 20 percent undecided, the trick is to figure out who is going to vote at all, and where the wavering voters will come to rest.

Pollsters say a key to predicting what undecided voters will do is what they did last time. Follow-up questions allow pollsters to cut the undecided number to 3 or 4 percent, Zemach said. In the end, they tend to "go home," voting the same way as before.

With the emergence of Kadima, that no longer works. Likud and Labor are traditionally Israel's largest parties, but they're suddenly also-rans.

"This is the hardest assumption we have to make," said Zemach, who has been conducting pre-election polls since 1973 and is Israel's best-known pollster, adding that "we no longer have the luxury" of putting voters back in their old parties.

All this makes Smith nervous. His own polls consistently show Kadima winning about 35 seats, but his gut feeling is different. "I don't have the impression that every third person is actually going to vote for Kadima," he said, warning of election night surprises.

Smith himself tripped up last November, when his exit poll showed elder statesman Shimon Peres winning the Labor Party primary. The winner by a wide margin was union boss Amir Peretz. Smith said there's a crack in the glass on top of his desk, where he slammed down his fist after figuring out where he went wrong.

In January, normally reliable Palestinian pollsters were way off in their exit polls, insisting that the ruling Fatah had won the Palestinian parliamentary election, when it was Hamas in a landslide.

Fears of similar scenario will be in play when Zemach and two of her colleagues go on competing television stations the moment voting ends on March 28 to present their exit poll projections.

Under the best of circumstances, it's a nerve-racking moment, Zemach said. This year, with the added variables, it's even worse. "I'm not going to do it again after this election," she pledged, because there's too much pressure, and that causes errors.

Gideon Rahat, a Hebrew University political scientist, has watched this with concern. He complained that the weekly polls, published on the front pages of the major newspapers and broadcast prominently on TV, have distorted the electoral process, turning it into a sport, with people picking favorites.

"The horse race aspect is problematical, because elections are not a horse race," he said. "They are supposed to be a decision between world views, ideologies and political programs."

Yaron Ezrachi, a Hebrew University professor specializing in culture and democracy, said the polls were "a form of disruption and diversion," but only one factor in an already imperfect system.