There has been some talk lately about suspicion that the polls are behaving strangely. Specifically, people dropping sharply in relation to others on the last day. I've been monitoring the poll data and making and analyzing charts, because thats what I do for fun ;D . There's actually an extremely simple explanation for exactly what is occuring, and lest more people become unnecessarily frustrated and angsty, I thought I'd elucidate the subject.
Before even looking at the data I hypothesized that on the last day people in the bottom six would frantically scramble to get above one another, sending the bottom and sixth-to-last position significantly up in votes. This is, in fact, what occurred.
28.9% of all votes were cast on the final 24 hours, so this allows for a great deal of movement. The bottomplace position jumped 25.7% between 10:00 EST and 1630 EST. During that same time period the sixth-to-last place jumped 21%, and fully 47% in the last 24 hours. To put this into perspective: the final position of sixth-to-last would have been 3rd place 24 hours previously.
What this means is you've gotta keep bringing in the votes or you are going to sink. More on that in a moment, but before that let me explain how the bottomplace people did that.
Method: But first let me explain briefly how I got my data. (feel free to skip this) I tried to cut paste the poll standings once an hour throughout the day. Cut pasting it into microsoft excel and then using sort functions pared this down to only about a minute of my effort each time, and I set up automatic calculations so based on cut pasting it it did the appropriate math and told me fun things like how many people voted total, since last iteration, where people moved in relation to one another, etc. I converted time into "Poll Hours", which, as you can probably guess, is how long the poll had been running at that point. It ran for 72 hours total, which is three 24 hour periods.
How to jump: So the question may occur to you, how did people scramble for votes and succeed? The answer is simply that asking your friends list for votes (or, yes, just posting the link in your lj with "there's a poll up" works just as well) does work. From my data this looks to be a solidly proven fact.
Most people post their link very shortly after the voting opens, which obscures the effect, but where I have identified people that have done it at later times, their jump in the standings has been dramatic and immediate.
HOWEVER, because this is an "artificial" injection of votes not related to the general trend of voting, persons who do this will see their position in the polls gradually go down as they are overtaken by people who either post later, or are simply doing better than them without friends list help.
Idolist A posted a link to the polls at Hour 44 and immediately shot up 10 votes in the next hour, from 36 to 46, bringing him/her from 32nd place to 24th. They eventually made 18th place before "only" gaining six votes of the last day's flurry and tanking down to the five way tie for 24th place which was just above elimination.
Idolist B posted about the polls some time Sunday morning, and shot from 36 votes at my last check Sat night (Hour 31.5) to 47 votes by Hour 39.1 (which means they got 34% of all votes cast during that time, since only 32 votes were cast overnight). That brought them from 26th place to 15th. They'd hit their highest position of 14th a few hours later, and stay around there for the remainder. The last day they got 10 votes, which seems to be what it took to hold position, ending at 16th.
One more example of a later-in-the-polls posting of link and then I'll do someone who posted immediately and just went down from there (D below). So feel free to skip C if it looks too much like B. I'm just including it because its me. (=
Idolist C is myself. I also posted a link later on in the cycle. There's pros and cons for doing this: Pro: you can see how you were doing without it, how much it effects things; Con: because people expect everyone to be following a trend by then they get freaked out and think you're being sheisty if you shoot up in the polls the second day. Its no different than the boost people get when they do so the first hour, its just more visible.
Anyway, the flip side to all this is that these "artificial" injections of votes effect everyone else. As noted if you do it in the beginning you'll see yourself slide in the end. Conversely, not doing it at first, I was "artificially" low at the start. I went from last at Hour 0.33, fourth-from-last at Hour 1, sixth-from-last at hour 1.33.... (frankly I think its also more satisfying to watch the polls in a manner in which you're constantly risign against everyone else rather than putting yourself at an unteniable position in the beginning and watching with dread as you steadily sink). By saturday morning I was pleased to find I had drifted overnight up to ninth-from-last. However, knowing from previous polls that the bottom of the chart would go completely haywire towards the end, I wanted to give myself some safety, and hey, everyone else is doing it right?
At Hour 20.5 I posted the invitation for my friends-list to visit the polls, read and vote. By Hour 22 I had 10 more votes (38), pulling me from 26th (ninth-to-last) to 14 place. I would hit my high water mark of 11th place at Hours 25-27, and then slowly drift downwards again (again the usual effect after affect of your "summon friends list" spell has worn off) to 16th place Sunday night. I would be fortunate enough to get 12 votes the last day and relatively hold my place (17th place down to 20th).
Idolist D posted the link to the polls minutes after the polls went up, and thus started out in 3rd at my first reading at Hour 0.33. Long story short (and I note this entry is getting pretty dang long), they followed an extremely steady rate of decline from there on out, ending up in 14th by Sunday night. They were then shocked to find the rug pulled out of them when they only got three new votes out of the 142 people who voted in the last 10 hours and plummeted to 24th place.
The explanation, of course, is that just as people posting in the first minutes got an "artificial" boost of votes, and those who did so in the middle did, so did those people in or near the elimination zone who posted or re-posted asking their friends list to go vote on the last day. This caused a major amount of change the last day, and if you didn't get votes from around 7% of those people (ie ten votes) you were going to lose position substantially in the polls.
The good news is that they did NOT merely vote for their friends. Myself and six of eight people I was tracking relatively held position (all without calling in the friends list during or near this time). alexpgp even gained ground, rising from third to first, and reassuring us that all is right in the world because his entry was excellent and clearly people voting the last day still took the time to vote for it (he gained 18 votes Monday).
So what does all this mean? It means first and foremost you should NOT be shocked when the polls change radically in the last 24 hours, that is to be expected. You should also EXPECT to be at your highest within 2 to 5 hours after making an entry linking to the polls and go down steadily from there thereafter.
It also means that yes, friends-lists matter, and "pimping" (as it apparently has come to be called by many) works.
Some might say this validates their "this is a popularity contest!" wails. I would say yes and no. Yes friends-lists matter, but people still vote for entries that truly deserve it and I notice my own friends list lets me know when my entry sucks by turning out in much lower numbers (I think acid rain was a perfectly good thing to write about hrmph! ;D ). And on any account, in my personal opinion, friends-lists should matter. This is LJ-Idol, and, as I've said "not just a writing contest" (and got jumped all over for and I'm sure will again). What that means to me is that, for example, friends lists are an aspect of LJ. And what makes you popular on LJ anyway? Its not that these "popular" people are the hottest (god knows I'm not) or have cool cars or something -- they have the most people that like to read their entries on a regular basis. They've gotten where they are through writing. So the relative impact of their friends list is kind of the aggregate bonus of months or years of good blogging and being a good citizen of the blogosphere.
I went and read the info for one of the LJ Idol spinoffs (::dodges Gary's shoe::) and decided not to have anything to do with it specifically because it WAS "only" a writing contest. Writing doesn't exist in a vacuum. I want to write where I have to worry about the demographic reading it (my friends-list I'm pretty sure is a totally different demographic than LJ Idol, so I have to carefully balance interest to both1) (and for example, the demographic of people voting in the last 24 hours seems to have had distinct tastes. Given entries by people they presumably didn't know they were much more apt to vote for some entries than others).
All that said, I do think the effect can and SHOULD be mitigated by frequent community only or contestant only (and the burden on Gary of doing that should be much less now that there's fewer of us) votes.
And while I'm on a soapbox, I agree with the sentiment expressed by a few others that there should be no more immunities and byes at this point. In the unlikely situation that I am given such a thing I pledge not to use it, and when placed on the jury I look forward to voting "none of the above" each time.
1 Specifically, the demographic of my friends list has signed on for my usual fare of impersonal entries filled with facts, discourse on history, and observations or essays about general things. The Idol demographic seems to really favour personal introspective entries which is something I almost never write. So I'm constantly challenged to make something historical or factual seem "personal" enough for the Idol crowd. This challenge is the real world and I love it.
LJ Idol Meta -- Voting Trends