That depends on the criteria, which could be "the candidate that most people accept/approve" (my personal preference), "the candidate that the most people will choose over all others", or something else.

IRV has failure cases against both of these, and I wrote a good example of an IRV election which explains the issue here: https://www.technologyreview.com/s/410622/scoring-the-candid...

Warren's page on this with a lot more details is here: https://rangevoting.org/rangeVirv.html