Opinion polls should technically predict election results quite accurately, but for reasons probably deep in human psychology they sometimes don't really get it right. Some bright spark in Iowa decided to create an options market to predict election results and other phenomena. This structure forces the people voicing an opinion to put their money where their mouths are and guess the result. And it is this system that is the most accurate in prediction these days. The market rules again! Naturally there are some obvious drawbacks (i.e. only those with money, access, risk-taking proclivities take part) but results are results. In New Zealand we have our own options market (https://www.ipredict.co.nz/) for the election and other issues. The tests differ in one just asks which one the respondent supports and the other, which result would you invest in.
Why should asking people what they actually feel not lead to the 'right' result though? Well, there can be wishful thinking; the person might want someone to win but not be inclined to go out and vote; the person could be tepid and go with a party that they have always gone with. Investment requires a bit more circumspection, objective considerations and reality checks (mind over heart). It eliminates those who aren't really that motivated.
This applies in real life too of course. One can profess one way, but whether they invest energy, consideration and action into things should be the prime test. The more accurate test. Actions do speak louder than words.