RCV123 online system handles ties among candidates facing elimination differently than most RCV systems. (Other than tiebreaking, we use the WIGM RCV system that is the standard counting method.)
We vary from the common tiebreaking method because in elections with thousands or hundreds of thousands of voters, ties are very unlikely. But our mission is to make RCV helpful to anyone who wants to make a group decision — including smaller groups with perhaps only 10 voters in a school club or 25 voters in a civic organization. In a small group election with five candidates and 20 voters, for example, there are very likely to be several ties as the rounds progress.
Conventionally, RCV uses random chance to settle any ties. We believe it would be unsatisfying for small voting groups to find that much of the outcome was determined by random chance.
So we developed a unique tiebreaking system that calculates a single number for each candidate based on their vote totals and the choice column they are in. The candidate with the highest tiebreaking number wins that tie. If that tiebreaker number winds up in a tie, then RCV123 resorts to random chance.
Each time a candidate is ranked first, that gets a much larger weighting than later ranks. Each first-choice vote is worth 1, and each subsequent choice 2nd-10th is worth 2/3 (.67) of the previous choice on a ballot. Ranks 11-20 are worth .90 of the previous round’s weighting. Then all the votes and weighting for each candidate in each column are totaled to determine an overall tiebreaker number. So in our method, for example, three 2nd place votes are worth very slightly more than two 1st place votes. But it would take 37 10th place votes to have the same weight as one 1st place vote. The results page for each election has a table that shows this calculation in detail.
Our tiebreaking method looks at all the choice data from every ballot. This is different from the rounds of counting — which only looks at the data from each round as it is calculated. For example, in actual rounds of counting, a candidate with zero first-choice votes will be eliminated right away, and any 2nd or 5th or 10th place votes they may have may not matter at all. But those 2nd, 5th or 10th place votes will affect that candidate’s tiebreaker number — and may help them win a tie against another candidate with zero first place votes.
If two candidates facing elimination have a tie, and have identical tiebreaker numbers, then RCV123 will use random chance to decide an ordered list of all the candidates. That “Tiebreaker of Last Resort” list can be found on the results page of any election.
The use of the mathematical tie breakers will be noted in election results with a blue rectangle over vote totals in that round for the candidates involved. The use of the last-resort, random tie breaker will be noted by the color green.
We believe our tiebreaking system is a good compromise between not weighting the choice column of votes at all, and excessively weighting one choice column vs. another immediately adjacent.