Abstract: Instant-runoff voting (IRV) is having a moment. More than a dozen American localities have adopted it over the last few years. So have two states. Up to four more states may vote on switching to IRV in the 2024 election. In light of this momentum, it’s imperative to know how well IRV performs in practice. In particular, how often does IRV elect the candidate whom a majority of voters prefer over every other candidate in a head-to-head matchup, that is, the Condorcet winner? To answer this question, this article both surveys the existing literature on American IRV elections and analyzes a new dataset of almost two hundred foreign IRV races. Both approaches lead to the same conclusion: In actual elections—as opposed to in arithmetical examples or in simulated races—IRV almost always elects the Condorcet winner. What’s more, a Condorcet winner almost always exists. These findings help allay the concern that candidates lacking majority support frequently prevail under IRV. The results also reveal an electorate more rational than many might think: voters whose preferences among candidates are, at least, coherent in virtually all cases.