Back some time ago, say around 1968 or so, when computers were just becoming available for less than a few thousand dollars (US or other) per minute, a series of articles in Scientific American magazine took a look at different methods of counting the votes, and at how different choices might affect the outcomes of elections.
One of the conclusions reached was that any time there are three or more viable candidates, the one elected will nearly always be the one considered UNACCEPTABLE by the maximum number of people. While there might be no more people happy with one of the other choices, there would always be fewer unhappy with any other choice.
Not all the calculations were shown, so it's difficult to say whether the result should be taken as gospel or just as prophecy (if there's a difference), but it has seemed to have at least some validity in elections I've observed from a distance. The series did suggest a minimum level of popularity required to call a fringe candidate real, and I haven't seen a US election where a third party candidate has achieved a "real" candidacy, but several other countries (some very nearby) have demonstrated some interesting samplings.
Maybe those results contained an element of validity.