In early October using tweaked issues that will be further adjusted in v1.1 coming out later, The Political Machine correctly predicted 47 of the 50 states in this year's election. No polling of candidates. All based on its issue model.
So why didn't it get all 50? The Political Machine's turn out model can't predict the actual turn out. It is based on historical norms. In this election, Republican turn-out was significantly higher than in any previously recorded election resulting in Bush victories in Ohio, Iowa, and New Mexico -- by narrow margins but still enough.
The professional polsters, by contrast were wrong by much more significant amounts in general despite having vast tracking polls and daily access to data we do not have.
The question one may ask is - will politicians in the future start making use of artificial intelligence, found in a primitive form in games like The Political Machine to determine where candidates should go and what issues they should speak on, and how much money to spend on advertising in that state.
When The Political Machine was being developed, political operatives like Hollywood Friend, Smear Merchants, and Intimidators and the like were thought to potentially be "unrealistic". And yet, as the election has shown, they seemed to be much more in-line with what the world saw. With hollywood celebrities doing special concerts, Swift Boat ads, and cries of voter intimdiation, even these units have their place.
The Political Machine will be updated in v1.1 so that it's not as specific to this election so that we can ensure that it remains fun long after the election is over. Meanwhile, players on both sides can crank up the difficulty level to the higher levels and see if they could have done a better job than Bush and Kerry at winning the election.