You're talking about completely different things.
Odds are an estimate of who will come out over the line quicker, polls are the underlying data used to calibrate that model.
The odds set in 2016 by various groups were often wildly bearish on Trump winning, the polls were not so much. Political polls and surveys are usually not off the mark by all that much, especially if you have good demography data.
Surveys can be manipulated but they are also much better than flying blind with no data at all.
