EPH & the GOP in the USA

So this post arose out of comments I made here.

Imagine if the US primary season and Presidential election was replaced by the Hugo voting process.

Everybody would get to nominate three candidates that they liked. The total number of nominations would be counted and the top three nominees would become the three finalists. There would be then a general vote in which people picked one of the three finalists (or No Award if they didn’t like the finalists).

Of course in this fantasy world elections would be very different but by the power of magic this change occurs overnight so that the US still has Democrats and Republicans and exactly the same pool of nominees as there are currently (OK maybe not Santorum and Gilmore because they are polling really low and Fiorina as well because 9 is an easier number to work with).

I’ll assume for the moment that also there are slightly more Republicans than Democrats and I’ll assume small numbers of independents and non-aligned people and people who’d nominate a mix of Republicans and Democrats. To make the numbers easy will say there are 100 Democrats nominating and 150 Republicans and 50 independents

The Democrats

The Democrats have three candidates Clinton, Sanders and O’Malley. O’Malley isn’t doing very well in the polls but for the nominating stage in our strange Hugo-verse America, that doesn’t matter. The Democrat votes need to nominate three candidates and they have three candidates. So Clinton, Sanders and O’Malley get 100 Nominations each.

The Republicans

The Republicans have a more varied field. The current state of play based on possibly useless polls looks like this:

  • Donald Trump 36.1%
  • Ted Cruz 16.3%
  • Marco Rubio 9.8%
  • Ben Carson 8.3%
  • Jeb Bush 5.5%
  • Rand Paul 2.8%
  • John Kasich 2.6%
  • Chris Christie 2.5%
  • Mike Huckabee 2.1%

But in our Hugo-verse America people have to nominate three people not pick their favorite. Trump is popular but divisive. Bush is unpopular but inoffensive etc. There are lots of ways of picking 3 things from 9 – specifically 9!/(3!(9-3)!)=362880/4320=84. This kind of race would be quite different to the current one. In reality Trump currently leads the polls but Trump is not necessarily popular with the supporters of other candidates. Possibly Marco Rubio would be the most popular second or third choice but there is no real way of knowing. However, what is quite possible is that no one candidate appears on nearly ballots. Rubio (or whoever) would need to appear on two-thirds of the all the Republican ballots to get 100 nominations.


Independents would add to the mix but in a very varied way, some nominating a mix of Democrats and Republicans and some just specific names or a third party candidate.


So in this Hugo-verse voting unless one Republican managed to get on two-thirds of all the Republican ballots, the odds are that ALL the nominees would be Democrats! Note that would even include Martin O’Malley who isn’t even that popular with Democrats!

Luckily in this Hugo-verse election all the Republican voters don’t have to vote Democrat, they can instead vote No Award. As Republicans outnumber Democrats in this fictional scenario the winner is No Award.

This is obviously a flawed system for nominating political candidates. While the pick-who-gets-the-most-nominations can work fine in other circumstances, it works badly in situations were a minority of voters have highly similar voting patterns. Obviously in the context of partisan party politics using a system like that would be crazy.

For Hugo categories it usually works OKish but it will work badly when there are slates or possibly even just different groups of fans with quite different interests.

How having an elimination process helps

In our bizarre Hugo-verse America the Presidential election has gone badly. Only Democrats got nominated and so the majority of Republicans have no candidate to vote for. How can we fix it?

The Republicans would have won a candidate if only there had been a smaller field of candidates. However, nobody wants to use opinion polls or multiple rounds of voting to sort out the viable from the non-viable candidates. EPH basically solves this problem by working as if there were multiple rounds of voting but without the voters having to keep filling in ballot papers.

After the first round of counting there has to be two candidates who have far fewer nominations than anybody else. At the first step the candidate with the fewest number of nominations is eliminated. For the sake of argument, let’s assume that was Huckabee.

Now here is the clever bit. Anybody who put ‘Huckabee’ on their ballot paper now has only TWO names left. When we come to the step of seeing which two candidates have the least number of votes the EPH algorithm counts those two names as having 1.5 votes. The missing vote for Huckabee gives the other two names a bit of extra oomph. Two new names are identified as having the least number of votes (including weights) and those two names are compared using the raw unweighted number of votes.

Now note because there are more Republican candidates for most of these rounds (which takes place in a fraction of a second in a computer algorithm) it is less popular Republicans being eliminated and so more and more Republican ballot papers have higher weights when it comes to elimination.

When the rounds come down to the most popular four or five candidates the Democrats may not have had ANY candidate eliminated yet. So most Democrat ballots still have the standard weighting. In this case Martin O’Malley and say Bernie Sanders (or maybe Hillary Clinton, depending on who gets the most votes from independents) face off. Then possibly Clinton versus Sanders. The pair comparison is always by raw votes but the pairs that have to face off is always with the WEIGHTED votes.

The net result would end up with at least one Republican (possibly two) and at least one Democrat. Note the algorithm doesn’t ‘detect’ Democrats or even know that Democrat votes were more likely to be very similar to each other.

It is a matter of design nor morality

OK, it might be morality as well but that is another argument. The point is the pick-who-gets-the-most-nominations works badly when there is a sizable minority of voters who vote in a similar way FOR ANY REASON AT ALL. The system breaks regardless of it is a slate or just a bunch of fans who happen to all have much less varied taste than other fans.

Now this is obvious when there is a slate in operation but it could even work more subtly just with different fandoms or pre-Hugo ‘buzz’ about particular works or Locus lists etc. Anything that makes some votes *more* similar will tend to give a boost to those votes in the current system. EPH doesn’t magically make that boost go away (because popular stuff will necessarily be popular with lots of people) but it lessens the impact.