Welcome to Jester's Trek.
I'm your host, Jester. I've been an EVE Online player for about six years. One of my four mains is Ripard Teg, pictured at left. Sadly, I've succumbed to "bittervet" disease, but I'm wandering the New Eden landscape (and from time to time, the MMO landscape) in search of a cure.
You can follow along, if you want...

Friday, February 10, 2012

Lies and damn lies

Posting frequency is down this week, and number-crunching is to blame.  ;-)  I've been spending a lot of time building the back-end behind three upcoming blog posts.  This is the first of those.

A few days ago, I indicated my belief that it's going to take about 3200 votes to be a top seven candidate in the CSM7 election.  Where did that number come from?  After all, in the CSM6 election, to get into the top seven only required 2240 votes.  So is Jester just talking out of his ass again?

Hang on tight, because I'm about to take you down the number-crunching rabbit-hole.

Let's start with this chart:


Stacked bar charts are often quite useful in comparative analysis, that is, the comparison of present-day results with like past results.  They allow you to do a bit deeper dive into the sources of those results.  You most often see them used in revenue analysis of large multi-national companies, or retail organizations with many stores or districts of stores.  How did store #2382 do compared to how it did last year?  Did it sell the same amount of electronics as last year?  How did district #4 do compared to district #5?  Did they sell the same total amount of electronics?

And more to the point here, how did the top store/district/whatever do compared to the top store/district/whatever last time?  That's what this first chart shows: how did the top nine vote-getters do in the first six CSM elections?  Specifically, this chart shows the percentage of the vote received by the top nine vote getters.  The chair of each CSM is the left-most bar, with the ninth-place finisher on the right.

What should jump out at you first when looking at this chart is that every bar is almost the same length.  What should jump out at you second is that every election except one (CSM4), the chair received nearly the exact same percentage of the vote.  Let's look at these in order.

When generating this chart, I didn't count every single vote.  Statistically, this turns out to be not all that useful, and it's not surprising.  Every year, there are more candidates or fewer, more interest in the election or less.  But every single election, there are votes placed for winning candidates and votes placed for losing candidates.  Recently, I repeated a stat that Trebor came up with, the stat that for CSM6, he said that 70% of the votes were placed for winning candidates (the actual number turns out to be 68.5%).  This entire blog post grew from the seed of me asking myself: "Is that unusual?"

Turns out, the answer is "no".  See how the CSM4 election is kind of an outlier on the chart above?  If you disregard that election, EVE players are very very consistent in this regard: 67% of CSM election votes are placed for winning candidates, election after election.  Based on this, I stopped counting losing votes and the analysis of the remaining data became much easier.  The chart above only counts votes cast for the top 14 vote-getters: the people that won either full CSM seats or alternate seats.  Once you do that, 80% of these "winning" votes are cast for the top nine candidates and 20% are cast for the remaining five, and again, this happens election after election.

That's why all those bars are the same length.

It also turns out EVE voters are very consistent when electing the CSM Chair.  Whether it's an election with a good turn-out or a poor turn-out, whether there are lots of people that care about the election or not, as you can see, five out of six times, the chair has received 16% of the winning vote.  Even more interesting, though there's some variance in the top five and the bottom eight, five out of six times, the seventh place finisher has received 6.5% of the winning vote.

There's another way of looking at the seventh place finish specifically: if you again take out the CSM4 election as an outlier, the seventh place finisher has needed between 16% and 47% more votes than the seventh place finisher in the previous election.

What does this say about CSM7?  Well, there are two ways of looking at it.

Taken as an annual measurement, each year, the turn-out for the CSM election has increased by 25% over the previous year.  This doesn't show very clearly if you look at each election separately, but CSM elections used to happen every six months.  If you remember that the CSM1 election took place in 2008 and the CSM6 election took place in 2011, the 25% growth curve comes out very clearly.  So, let's assume for a moment that there's a 25% increase in total voters from CSM6 to CSM7.  That's 61,370 votes for the CSM7 election of which 41,000 or so will be votes for winners.

If that's the case, the Chair (whomever that is) will take 16% of that: 6375 votes and the seventh place finisher will take 6.5% of that: 2675 votes.  So, if voter turn-out again increases by 25% this year, then to ensure a seventh place finish in the CSM7 election, you need 2675 votes.

But let's assume for a second that The Mittani is going to be the Chair.  Last year, Mittens tweeted that he received 1700 Goon votes.  I believe that to be true.  He also received 3600 or so non-Goon votes.  Last year, the Goons officially supported two candidates, but actually split their vote in three because there were two unsupported Goon candidates that nevertheless drew off Goon votes.  GSF is also bigger this year than it was last year, and has a large pet contingent.  Let's assume that this split does not happen this year and Mittens picks up 500 or so votes from those pets.  That will give Mittens between 3500 and 4000 Goon votes.

Last year, I estimated that Mittens received about 1500 of what I called :lolcsm: votes... votes from people who hoped that Mittens would just create some funny e-drama for them.  These are the voters that put people like Mazzilliu on the CSM (where she delivered on that implied promise).  Sadly for them, Mittens proved himself a non-dramatic, non-funny, reliable CSM Chair.  As a result, I don't think Mittens is going to receive those :lolcsm: votes this year.  They'll instead go to people like Mintrolio and Xenuria.

But I also estimated he received between 2000 and 2500 votes from people who were voting for him as a serious candidate, or because of name recognition.  There's every reason to believe that Mittens will actually pick up more such votes.  I'm inclined to think, in fact, that he'll double or nearly double this count of "serious" votes.  All in all, I expect Mittens will pick up close to 7500 votes total in the CSM7 election.

If he does, and that again turns out to be 16% of the total "winning" vote, then the total vote count will be 72,200 votes in the CSM7 election, of which 48,300 will go to winning candidates.  That means to ensure a seventh place finish in the CSM7 election, you need 3150 votes.

2675 votes represents a 20% increase from what was needed for that place in CSM6.  3150 represents a 40% increase.  Both are well within the historical range of 16% to 47% that I mentioned earlier for a seventh place finish.  So, consider that the low- and high-end for seventh place as far as I'm concerned: between 2675 and 3150 votes.

And if you're running for one of those seven seats, I'd hope for the lower number, but plan for the higher.

Six weeks and one day to go in the CSM7 elections.

14 comments:

  1. I'm firmly in the :lolcsm: camp, have never voted in the CSM election and won't this year. Unless I can figure out a way to sell my vote. Have there ever been organized vote-selling or vote-buying campaigns? Is such a thing allowed by CCP?

    ReplyDelete
    Replies
    1. No rule against it, but there's no easy way to prove who you voted for so it would be a challenge to buy votes in bulk.

      Delete
    2. I wish votes were an ingame item.

      Each account can redeem one, they can be traded, transported, grabbed from loot and contracted just like normal items.

      At the end of the election, counting is determined by the largest stack of vote in each candidate account (if a player has more than one stack, only the largest count).

      Maybe this would even increase the vote count, since it voting would be done in the client.

      Delete
  2. How is it not allowed, unless ya use RL money. Personally I am a fence pole sitter between :lolcsm: and one dude with some decent ideas. Dont know how I will vote though still.

    ReplyDelete
  3. That is a fascinating analysis, Jester. Do you feel it is unusual or remarkable that CSM elections results are so similar year over year, in terms of percentage of votes cast for the top 9 seats? Any theories as to why there are such similar results, or is that for another blog post?

    ReplyDelete
    Replies
    1. The more I think about it, the more I think, no, that 67% of votes cast for winning candidates is likely to stay pretty consistent. I think there's two human tendencies at work here:
      1) the tendency for those voting for winning candidates to want to "pick a winner" and not expend votes on candidates with little shot; and,
      2) the tendency for those voting for losing candidates to want to fight on for a friend despite all odds. ;-)

      But I have no hard data to back this up, and thus, I don't feel like it rates its own post.

      Delete
    2. There may be one exception to the above...the summer of rage and its resulting chain of effects thereafter. This all resulted in Crucible which may well increase/decrease the consistency shown thus far.

      Just an observation, not an argument to what is otherwise pretty sound numbers. Good Job! :-)

      Delete
  4. I agree with most of your analysis except for one (pretty important) part.

    Your are qualitatively analyzing/upgrading The Mittani's vote count and - based upon the unrelated quantitative relationship between first/total vote count - estimating total voter turnout.

    (Mind, you actually used two unrelated methods of estimating an increased voter turnout, simply extrapolating from last elections and the just described upscaling of the qualitative FirstSpot-increased-vote argument), without explicitly noting the difference.

    I think the extrapolation method one is valid: "votes have always increased, they will continue to do so.".

    But the other isn't. Your qualitative analysis is very much based upon increased appeal and voter-base of one candidate compared to the others. That EATS INTO their vote-share. You didn't make an argument that The Mittani's "popularity" will substantially increase overall turnout. That aside, mixing quantitative and qualitative analysis on voting patterns is always tricky.

    My predictions:
    ---------------
    Qualitative/total vote count related: Subscriber number hasn't increased, it might even have decreased. Some of that might be compensated by higher turnout, because the CSM has been perceived as having much more of an impact than in some past cycles. We will have increased numbers, but maybe less so than in the past.

    The relationships between individual slots will indeed be similar again, but your arguments ex. The Mittani have actually convinced me that he will be getting more than those usual 16% (slightly depressing the vote of the following candidates).
    That has no impact on the question of total voter numbers. Just wanted to mention that again. ;)

    Like your blog,
    keep doing the good work.

    ReplyDelete
  5. Since you're doing statistics (lies and more damned lies), have you considered testing the various electoral results against Benford's Law? Ofcourse, don't use a Chi Squared test, use a Kolmogorov-Smirnov test, or a Kuiper test instead.

    ReplyDelete
  6. You make (at least) three assumptions:

    1. 67% of all votes will be cast on winning candidates and each candidate will receive a fixed % of it.

    2. There will be a 25% increase in total voters.

    3. The Mittani will get a lot more votes, from existing voters.

    These assumptions cannot be true at the same time. Your first number (2675) comes from the 1. & 2. assumptions. When you add the 3., for some reason you choose to keep the 1. In this case the total increase in voters should be ~47%. I think it’s more logical to assume that the percentages between candidates change, since those voters won’t vote for someone else and not enough new voters to compensate (if 2. & 3. are true). That means other candidates get less votes, therefore you need less votes to get in (if every candidate lose the same % of voters, excluding the first place of course). I had to add this 4. assumption, because the 1. is not true in this scenario and I can’t say anything about the last place without it. In this case The Mittani will receive ~18% of the winning votes and… I don’t have the data to calculate the last place number. It should be a tiny bit under the original prediction.

    Now about those assumptions, going backwards:

    The third one is based on pure speculation.

    The second one has some data behind it, but the increase is clearly unsustainable in the long term. (See: http://xkcd.com/1007/ ) It requires 25% growth in the voter base, and thus sooner or later in the player base. That amount is simply unrealistic. Currently we are far from 100% turnout, so the increase can be 25% or more.

    The first one is based on interesting data. My only problem with it is the exclusion of the CSM4 data. Ignoring data that is not consistent with your assumption is pretty much cherry picking. I think the data is interesting enough even with the CSM4 results.

    Nice work and insane posting frequency. :)

    ReplyDelete
  7. This supports my theory on why you have Goon/Test candidates like Lyris and the guy who runs Dreddut running now. They may end up canabalizing votes from Mittens but they are still ensuring the distribution of votes are going to them; nullsec bloc and taking votes from the non-nullsec bloc

    ReplyDelete
  8. LIMES, LIMES AN DANMED SETISTICKS.

    CONFRIMIGN TOO MANY NUMBRE CONFUSE VOTERS. ALSO VOTE FUR MINTROLIO TO END ALLES CSM SETISTICKS DIMSCUSIONS.

    KEEP UP GOOD BLOGGIGN !

    ReplyDelete
  9. I wish that players had the option of either a "vote for" or a "vote against". In such a system, the candidates from the larger alliances might not have an automatic lock on a position if disparate players all utilized their "vote against" option to block a candidate that had shown themselves to be abusive.

    ReplyDelete
  10. Interesting, number of likes received as of feb-14-2012 1200:
    The Mittani - 1181
    Dovinian - 595
    Two step - 442
    Seleene - 427
    Hans Jagerblitzen - 359
    Kelduum Revaan - 342
    Meissa Anunthiel - 314
    Elise Randolph - 306
    Greene Lee - 295
    riverini - 280
    Fon Revedhort - 265
    Trebor Daehdoow - 190
    Mintrolio - 190
    Prometheus Exenthal - 184
    Alekseyev Karrde - 184
    Draco Llasa - 179
    Sollana - 155
    Mike Azariah - 150
    (plus seven others between 100 and 149)

    Should be interesting to see if the number of likes at the end of the nomination is any indication of CSM final positions.

    ReplyDelete

Note: Only a member of this blog may post a comment.