Ranked choice voting, specifically instant-runoff voting (IRV), has many advocates and has recently been implemented in various localities. While this is generally preferable to first-past-the-post (FPP), IRV has serious flaws. Not only is it possible, but likely, that the consensus candidate will be eliminated in the early rounds. When multiple candidates are relatively close to each other the candidate in the middle is effectively squeezed out.

Consider this visualization of a political map. There are four candidates represented by different colors. The map is filled with the color of the closest candidate. Given an even distribution of voters, Yellow is at the center of public opinion.

Using IRV, Yellow would be eliminated in the first round because they were ranked first by the fewest voters. Red ends up winning in the third round. This method of counting considers the voter's preferences but fails to capture the strength of support for each candidate.

A Score ballot, in addition to capturing preference order, captures each voters level of support for each candidate. This additional data is useful, but simply adding up the scores ignores the preference order and can introduce unintended strategic incentives.

Adding a runoff round can reduce strategic voting by encouraging voters to differentiate support between candidates. This variant of Score voting is called Score Then Automatic Runoff (STAR) Voting. STAR Voting first calculates the quantity of support from the entire constituency to select the most viable candidates. Each voter's full vote is then applied to the viable candidate they scored higher.

The results from 5,000 simulated elections with 1,000 voters and 3 randomly generated candidates show that STAR voting is more effective than IRV at selecting the candidate closest to the center of public opinion. IRV gets further from the consensus candidate as the number of candidates increase, while STAR remains stable. These simulations do not account for things like strategic voting but given the failure rate under ideal conditions (conceptually, voters with perfect information voting honestly) suggests fundamental defects with IRV.

The following graph illustrates the behavior of FPP, IRV, and STAR Voting. The baseline is the position of the candidate closest to the center of public opinion. Random, a plurality voting system in which voters randomly select a candidate, is included for reference.

The following series presents the same data with more detail. The center of each box represents the average distance from the consensus candidate. The sides of the box are one standard deviation from the average. The raw values are displayed below the chart.