"We're not making up the drama of this situation," Hermann Groehe, Merkel's campaign manager, told a small group of foreign journalists when asked about unreliable polls.
"It's not a screenplay we created as part of some campaign strategy. It's the real situation, and it's going to be a close race right to the wire. It's all still very much wide open."
(Read more: Germany probes breach after half-naked man boards Merkel's jet)
A trawl through data shows pollsters over-estimated the strength of the CDU/CSU in every federal election since 1990.
Merkel's conservatives may have a comfortable 15-point lead over the next biggest party, the Social Democrats (SPD), but her center-right coalition is in a dead heat against three combined left-of-center parties in some opinion polls, each having about 45 percent.
She knows that in 2005 nearly a third of voters made up their minds in the final week.
"This will be by far the most difficult German election to predict ever," said Wolfgang Gibowski, a political scientist at the University of Potsdam, who co-founded the Electoral Research Group (FGW) polling institute in 1974.
"It's a huge mistake to think 'Merkel's got it sewn up'. There's a lot of volatility, and the pool of swing voters is larger than ever before. Party programmes are so similar, so it's easy to see voters changing their minds up to election day."
That gives sleepless nights to the pollsters, who blame a fickle electorate for the waning reliability of their polls.
(Read more: German elections are a 'close call': Merkel)
"Our work has become harder because voters don't have the same close ties to the parties any more, and before there were just three main parties compared to six or more now," said Manfred Guellner, managing director of pollster Forsa.
His institute was nevertheless more accurate than rivals in predicting a fatal slide for the CDU/CSU in 2002 that cost Edmund Stoiber victory. Forsa was also the only pollster in 2009 that did not over-estimate CDU/CSU support.
The conservatives' faith in polls took another beating in January when they lost Lower Saxony state to the SPD and Greens. The CDU was forecast to win 41 percent but won 36 percent.
"Voter turnout keeps falling, and that amplifies the impact of these last-minute shifts," said political scientist Thomas Jaeger at Cologne University, predicting turnout could dip below 70 percent from 70.8 percent in 2009 and 77.7 percent in 2005.
((Read more: Is Germany in 'cloudcuckoo land' over Greek debt?)
Another source of uncertainty are two new protest parties on the ballot:both the Pirates, which campaigns for e-democracy and citizens' rights in the info-technology sphere, and the eurosceptic Alternative fuer Deutschland (AfD) poll 3 percent, but may do better than expected.
Due to an intangible embarrassment factor associated with the AfD's unwanted popularity with the far right, which is ostracized in Germany, the party could get votes from people who are less than candid when talking to pollsters.
"The problem with these smaller parties is that supporters won't always admit that to pollsters," said Jaeger in Cologne - adding that this also applied to the Free Democrats (FDP), the junior partners in Merkel's coalition.
The FDP have repeatedly confounded dire poll predictions. They were not expected to survive the Lower Saxony election but won almost twice the level of votes forecast.
(Read more: Germany's Merkel moonlights as history teacher)
Peer Steinbrueck, the SPD candidate to unseat Merkel, hopes for a late swing that will make pollsters look foolish, telling reporters: "It's like a football match - it is the final 10 minutes that count."
Guellner at Forsa says the SPD's success in tapping into the dormant pool of millions of former SPD voters who have stayed at home in the last two elections is a major factor of uncertainty.
If the SPD mobilizes an extra 2 million voters it would reduce the CDU/CSU's share to about 37 percent from the current 40 percent. That could thwart Merkel's ambitions and force her to repeat the 2005-2009 right-left 'grand coalition' - or even hand the SPD and Greens a long-shot victory.
"That's something that the pollsters will never be able to definitively answer - how many people will stay at home on election day and how the undecideds will behave," said Gero Neugebauer, political scientist at Berlin's Free University.