SOME polls were accurate.
The polls by professional polling houses (Gallup, Rasmussen, ARG, SurveyUSA, etc.) showed a steady Obama lead for a month leading into the election. They were largely correct. MSM polls showed Obama with unsupportably huge leads (10-15%, when he actually only won by 7. These usually overstated Obama's support, and for much the same reason as they are today - messed up internal demographics and unrealistic partisan ID breakdowns.
The RCP average was very close to the actual result.