Scoobie, the relationship is that all those electoral maps are derived from the polling data. Each of them is based on some sort of synthesis of multiple individual state poll results (the mix varies according to where you get the electoral map). The electoral maps take in some no-brainers, nobody expects Obama to lose CA or IL for instance, nor McCain to lose AZ.
The maps do tend to make things look much more set than they really are. They are after all entirely dependent on blended polling results, and therefore carry forward any systemic bias in the polls they use. The mapmakers blend the poll results to try to cancel out bias and errors, but if there is a fundamental common defect in the polling strategy/methodology of multiple polling sources, that will be carried forward into the map-making, possibly even magnified in it. They also lag the day-to-day state of public opinion by probably at least 72 hours.
So bottom line, the poll results are derivatives of real opinion taken through the filter of the polling methodology, and the maps are further derivatives of the polling results. Recent election results (past 10-15 years) have shown the polls to be far from perfect, and not particularly precise predictors of the final result until the last 72-48 hours before the vote. Of course those more-precise final results come out too late to get cranked into the maps, and may not even make it into the public eye before the real vote.
