Still confused? Consider these scenarios.

Intrigued, yet a little puzzled, by the concern commentators Cilek and Marchetti raise above about instant-runoff voting potentially causing voters to undermine their preferred candidate by casting their first place votes for him or her, I worked through some scenarios.

Here is a hypothetical ranked-choice election, where the problem seems to arise:

Imagine an election with 100 voters and three candidates: Tom, Dick, and Harry.

The voters cast their first-place votes this way:

Tom: 43

Dick: 29

Harry: 28

Dick's voters:

Tom: 6

Harry: 23

Harry's voters:

Tom: 8

So what happens? Under IRV, Harry, with the fewest first-place votes, drops out. His voters' second-place votes (8 for Tom, 20 for Dick) are distributed. And the final result is:

Tom = 51

Dick = 49

Tom wins!

But now imagine an alternative outcome. Two of Dick's voters (among the six who had previously placed Tom second) change their minds in the voting booth and decide to support Tom with their first-place votes. The initial outcome now is:

Dick: 27

Harry: 28

This time it's Dick who has the fewest first-place votes and drops out. His voters' second-place votes now look like this:

Tom: 4

Harry: 23

Tom: 49

Harry: 51

Harry wins!

So by voting for Tom in first place rather than in second place, two voters have cost Tom an election he otherwise would have won.

IRV proponents say such outcomes are highly unlikely, and that the same ambiguities affect any election (including a primary) that narrows a field of candidates for a runoff vote. But in theory it does appear that votes could backfire under IRV.