When an election gets more than 10 nominations, primaries are held to select the top candidates. Currently, primaries select the top 10 nominations so that no election will have you choosing among more than 10 candidates, but that can be awkward when you are asked to pick the top 10 candidates... out of 11. Spending three days throwing out one candidate sound both awkward and a waste of time.

I think it would make more sense if primaries were still triggered upon reaching 11 nominations, but then they should pick the top (say) 8 candidates. This would make primaries always meaningful (cutting down on the amount of choices to the most prominent/popular choices) but never awkward (you'd always prune at least a few people, and not just one or two).

As proof, I ran the elections of Stack Overflow, Programmers, Super User and Gaming, with and without the users that came up 9th and 10th at the primaries. The differences are negligible.


The full ordering of the candidates can be obtained by running the elections with 1, 2, ..., 10 seats. The 4th place candidate is the one that wasn't elected with 3 seats but is elected with 4.

To simulate what would happen when the 9th and 10th users in the primaries (marked yellow in the tables) would have been excluded in the elections, I used the "Withdraw candidates" option of OpenSTV, so that any preference for the two users is discarded correctly and the relevant ballots are updated correctly.


In no cases was the ordering of the remaining candidates changed by this, and only 50 votes from Stack Overflow (out of 4,269) were completely removed as the result of ignoring the bottom two candidates. I think it is fair to conclude that running an election with 8 or 10 candidates makes no practical difference.

Data porn





  • Also consider the alternative where you need 13 candidates in order for the primary to kick in, oh wait, people don't like that number. I'm just saying that picking 3 candidates out of 8 instead of 10 might have side effects. Another interesting statistic to consider is to see whether someone that would fall out under your system has become a moderator, because you could then be pruning out possible great candidates. Visible up/down votes are often biased, while anonymous 1/2/3rd choice votes actually make the real choice... Commented Aug 13, 2011 at 12:15
  • 1
    In the end there's only room for 2 or 3, that means at least 7 'possible great candidates' are going to lose out to the great candidates anyway. I think badp's proposals is a nice way of not hurting anyone's feeling, but they don't require any more attention than that.
    – Ivo Flipse
    Commented Aug 13, 2011 at 12:25
  • @Tom Yes, I understand the risk of losing out on candidates, and I'm hacking together something to check that in a sanely way. No promises though.
    – badp
    Commented Aug 13, 2011 at 12:26
  • @TomWijsman The same thing could happen with the current system
    – Wipqozn
    Commented Aug 13, 2011 at 12:26
  • 1
    @Tom The same thing meaning losing great candidates due to the bias of up/down votes. That bias exists regardless of how many people you eliminate.
    – Wipqozn
    Commented Aug 13, 2011 at 12:30
  • @Wipqozn: What? You are repeating yourself. The bias gets worse when you narrow the limits, I've not stated that it should not exist. -- The question we have to ask ourselves is "Why should visible primaries voting have more effect on the actual moderator election?"; you can't blindly implement changes like this and say that they improve the voting system. Commented Aug 13, 2011 at 12:31
  • 2
    What bad π is saying, I think, is that if there are more than 10 candidates lose at least 3 to get the numbers down. So when there are 11 you get 8 going through, 12 gets 9 and 13 or more gets 10.
    – ChrisF Mod
    Commented Aug 13, 2011 at 15:38
  • @Chris No, I mean for simplicity that it should be 8, and not 10 candidates going through.
    – badp
    Commented Aug 13, 2011 at 17:03

1 Answer 1


It always sucks to just meet the threshold and require a primary to eliminate even just a single person, but that’s exactly the purpose of the primaries. We want to narrow the possible of choices going into the election even if it is just by one.

Spending three days throwing out one candidate sound both awkward and a waste of time.

Notice, the total duration of the election didn't change at all even though a primary phase was introduced. The primary phase was always accounted for in the total duration of the election - not having a primary would mean more time would be spent in the nomination phase.

  • 6
    I'm not saying primaries aren't useful, I'm saying primaries would be more useful if they selected the top eight users. As I attempted to prove above, this does not really change the results of the elections.
    – badp
    Commented Aug 13, 2011 at 16:01
  • @bad as you have shown, elimination of people in the primary has no statistical significance in the final outcome, however, what you and I can't measure is if having a lower number of candidates in the final election impacts voter turnout in a negative way which has historically been a problem. We want people to be encouraged to participate and vote even if their choices would only be the 9th and 10th placed candidates. Commented Aug 13, 2011 at 17:36

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .