12-08-10 – Optimus Prime

orionid: Very interesting. I suspect you’re right. Jack and Suzy see a theme they’ve got the perfect picture for and enter it, then let their friends and family know about it, who then end up voting for others as well. They then disappear into the sunset until the next perfect theme comes along. Or as the data seems to indicate, most of them disappear forever. I’ve got 5596 unique entrants in my database and only 411 of those have entered more than 10 contests and 3211 (57%) have entered only one.

interesting supposition but it presupposes that Jack & Suzy’s assorted friends and family have fark accounts. in my personal experience my family (a sister) doesn’t. she merely looks because i send her a link. OTOH among my few RL and online friends 3 have fark accounts. and all enter at least occasionally. they lack my enthusiasm for entering regularly but match your “oh i have a picture for this” group.

basically i agree with you and orionid that some themes bring out more casual entries and therefore votes but wonder how many extra non participating voters that produces. in other words do Jack and Suzy’s family/friends really contribute significantly to the voting totals.

and since i don’t (usually) vote for mine what does a 4 vote total tell me about a particular entry 🙂

3 – Why is envotiation still going?

So I can get 8 more votes from the really late-comers and make my average.

I’ve only got this level of data for the last 7 contests, but 5 of those closed within 7:45-8:15, one at 9:15 and then this one at 4:30pm (all times in Fark server time).

Basically, I figured the overall vote drive is driven by the number of casuals, and the way to find any truth in that would be by creating an algorithm as such…

Edited to add: I screwed up the calculation of casual entrants in this post, so please disregard all statements about casual entrant counts and relationships to votes. All figures quoted as casual entrant counts are actually regular entrant counts.

Dangit, orionid, you got my interest up so I had to go run some numbers. Turns out the data doesn’t support this theory, at least not strongly.

Those with mathphobia may want to skip to the next post…

I only used data from the last 3 years (1/2/2008 – 12/8/2010) to exclude some of the wilder swings seen during the frontier times. I used your criterion for “casual entrant” – someone who’s entering this contest but has entered less than 8 of the previous 15.

I ran a correlation calculation on the various variables against the vote counts. (For those not familiar with correlation, it’s a measure of how closely one dataset is tied to another. A correlation of 1 means the datasets are tightly bound together – a 10% rise in one variable should yield a 10% rise in the other. A correlation of -1 is a perfect inverse relationship where the second variable falls when the first rises.)

The strongest correlations are those that are intuitively obvious. The correlation of total number of entrants to total contest votes is 0.87, total entries to total votes is 0.91 and total entrants to total entries is 0.96. You can see that relationship in this graph:

But the correlation of casual entrants to votes is only 0.60. You can see in this graph that the vote count swings wildly even with the same casual entrant count:

Surprisingly, the correlation of casual entrants to number of entries is even weaker at 0.57. It seems to suggest that the theme itself has more to do with it as it encourages more entries from even the regulars and more entries equals more votes.

Those with mathphobia may want to skip to the next post…

…Dude~~that was harsh! But I got a nice buzz goin’ on…

I ran a correlation calculation on the various variables against the vote counts. (For those not familiar with correlation, it’s a measure of how closely one dataset is tied to another. A correlation of 1 means the datasets are tightly bound together – a 10% rise in one variable should yield a 10% rise in the other. A correlation of -1 is a perfect inverse relationship where the second variable falls when the first rises.)

//Just repeating with different spacing; in the original post the negative sign got separated from the second 1 and it was confuzusing. 😀 Lord help the mathly inepted.

The voting graph is very interesting. Mostly, it’s an initial rise, then a period of steady-held interest, then a big plateau with a major climax, then a sudden fail. From there it has to recharge until something perks interest again.

//That’s what she said.

Recently, it’s been up and down, up and down. It’s either hit or miss. I wonder why the change?

you can’t really read anything from the shape of that graph as it’s sorted not chronologically, but first by casual entrant count and then by votes, which gives the drop-to-minimum, rise-to-maximum shape for each casual entrant count.

Oh. 😕 So there’s no real trend there. 🙁 Hmph. You just had to go and kill my buzz… 😆

1- Hey Kashari – these made me think of you – http://photo.net/photodb/member-photos?photo_id=11769932

Wow, those are awesome, thanks for the link U-Man! I wish I had more frequent access to horses out in scenic places. Or, even better, that I had my own again…Kashari was the name of my Appaloosa I had when I was younger.

btw…I LOVED your kitty pics that you linked. What did you use as your background, do you have a studio setup kind of thing?

I kind of have to stifle a laugh at the hypothesizing going on about voting correlation re: casual participants. To me it seemed perfectly obvious that themes with higher numbers of casual participants will garner higher numbers of total votes because if the theme generates more casual interest participants it will generate more casual interest viewers as well.

Hey, your husband started it! 🙂

But I screwed up in my previous post so it turns out you’re mostly right (as indeed should’ve been obvious). The figures I used as casual entrants are actually the regular entrant counts – those that entered 8 or more of the past 15 contests. So, the correlation between regular entrants and total votes is 0.60. The correlation between total casual entrants and total votes is 0.82, a much stronger relationship.

My suspicion is that the threads with a higher ratio of casuals to regulars than the average thread are the ones that coincide with higher vote totals.

I neglected this part of your hypothesis so I ran those numbers (and this time I think I’ve got it correct). It turns out that although the total count of casual entrants affects the total vote count somewhat strongly, the actual percentage of casual entrants does so less strongly. The correlation of percentage of casual entrants to total votes is 0.54.

I think the only thing we can safely deduce from any of this is that more entrants equals more entries which equals more votes. Which is kind of a “no duh!” statement.

But orionid‘s other idea about a “wall of win” got me thinking that there are two ways to measure the “success” of contest – total votes, sure, but also votes/entry. I think votes/entry may better indicate viewer interest as opposed to participation rate.

So I ran another set of correlation calculations for votes/entry and basically there’s nothing really to predict that (aside from the theme itself). The number of entries has absolutely no correlation to votes/entry (0.01). The strongest correlation is for percentage of casual entrants, but it’s a weak negative relationship at -0.27! So, a higher percentage of casual entrants in general somewhat equals a lower vote/entry.

Although December is always a huge fall-off month for participation and voting so no need to panic over this week’s total votes imho.

You’re right there (again). December not only produces a lower total vote count:

but it also produces a much lower average vote/entry:

And finally, for those of you still reading this, as kestrana suggested, here are the top ten themes that garnered the most votes over the past 3 years:

Date        Theme                                   Votes





2008-07-30  I Tawt I Taw A Puddytat                  5546

2008-06-04  Trees                                    4092

2010-01-20  Sunrise. Sunset.                         3691

2008-03-19  Nature Photography                       3559

2008-07-23  Farktography Classic: Blue II            3282

2010-09-22  Farktography Classic: Green 2            3067

2010-10-20  Travel Photography                       3047

2009-10-07  Farktography Classic: Black and White 2  3031

2010-01-06  Animals                                  3024

2009-05-27  Birds                                    2955

and here are the top ten in terms of viewer interest as indicated by votes/entry:

Date       Theme                                             Votes/Entry





2009-06-10 Tilt-Shift Faked Miniature Scenes                       17.38

2009-03-18 Tableau                                                 15.44

2009-08-26 Training Documents                                      15.41

2009-12-02 Win, Fail, or "I Like Where This Thread is Going"       15.10

2010-12-01 The Song Remains The Same                               14.80

2010-01-20 Sunrise. Sunset.                                        14.20

2010-09-22 Farktography Classic: Green 2                           14.20

2008-07-23 Farktography Classic: Blue II                           14.15

2009-01-21 Take Three                                              14.14

2009-04-29 Farkstronomy                                            13.80