Tumgik
#the most detailed
sevdrag · 1 year
Text
spoilers: i am an engineer
hello tumblr, and welcome to the most well informed shitpost you're going to see today: some Fun With Data involving the Harry Du Bois vs Vriska poll that just went around.
We started watching the poll just as the swing in votes started to happen, and as such, we're missing some crucial data I'll be upfront about. The poll started at noon, and we didn't start taking in information until the percentages started to change, at 13:53, so we don't have information on total vote count for the first 2 hours of the poll. And in the rush to nerd out about this, our first few data points of the vote swing - where you see the most change - didn't have total vote count either. There is also a brief period where I had to take my cat to the emergency room, and a longer period where my ass slept because this is just a silly tumble poll. Those present as straight lines in charts like this, but keep in mind, they could have had wild variation for all I know. However, based on the results, I doubt it. Big thanks to @smolalienbee and @pass-the-salt and I think @orderlyhouse for screencapping this data and relentlessly spamming the discord with it so that I could put it in this spreadsheet like the nerd I am. (do YOU have screencaps? DM me!)
Tumblr media
Early on in polls like this when total vote count is low, you expect to see big swings -- with 4000 votes total, for example, the amount you need to swing a vote 1% is 80 people. As these go up, however, obviously, the amount of votes needed to swing 1% increases. By the end of this poll, 1541 tumblr users would have had to make UNCOUNTERED votes, that were NOT cancelled out, to move the poll 1%.
According to our data, the swing you see above took place over a two-and-a-half hour period. For the remaining 20 hours of the poll, votes one way or the other weren't significant enough to be seen in the tumblr results.
For those who are interested, our first data point showing the 50/50 split hits at 16:21, when total votes were at 14194. From this point on, voters would have needed at least 248 UNCOUNTERED/NOT CANCELLED OUT votes to change the poll a single 1%. Just numbers to keep in mind! We talk about this at the end :)
Any winning margin smaller than 248 would, therefore, not register to tumblr. Keep in mind that by the end of the poll this number was 1541 votes required to swing a percentage, because the total vote pool grew so heavily.
In the end, as it turns out, after 77,019 votes, the poll was won with a margin of only 618 votes. final percentages were, surprisingly neatly, 50.4% v 49.6%. Won by 0.8%! Now that's a close poll.
(thanks to @mio-nika for grabbing these and @smolalienbee for confirming!)
Because I was additionally curious, and a huge nerd, I wanted to go in and calculate votes per minute to see whether there was anything weird going on. I am going to be polite and put this under a cut to save your dash from my utter nerd joy, but I have caveats to put in here first.
The first thing to keep in mind is that your data is only as good as your collection method! In our case, it was a very unprofessional method that went: someone took a screencap and posted it in Discord, and I entered it into a spreadsheet with the time entered as whatever time Discord registered it. While this ended up being pretty accurate for a 24-hour poll, it presents the problem that will come up when we start talking about votes per minute: I only could measure in increments of a minute, because getting into seconds here would have been (a) impossible and (b) super overkill. However, that means that there's more noise in the votes-per-minute data because everything is assumed to be in increments of 1 full minute. Good scientists acknowledge the weaknesses of their data sets as they do their analysis! (Source: I worked in research for 15 fucking years lol.)
Here is my ESTIMATE of votes per minute at each of the data points taken above. The green trendline is a rolling average that tells the story of the poll, more or less: Lots of quick voting at the beginning, calming down over the (American) evening, then ramping back up at the end. Much like we'd expect, I think.
(And again, remember that the bit in the middle where Americans AND Europeans were asleep presents as a straight line here but could be ALL over the place. We don't have that data because we were sleep like good feral grownups.)
Tumblr media
The noise at the beginning, I suspect, is again because I'm measuring over small and discrete increments of time. However, unless Tumblr wants to help me develop an app dumping real-time poll results into my EXCEL 24/7, there's no reasonable way I can see to get around this. Because of the noise there's a big standard deviation on here that I'm too lazy to really dig into because I did all of this while I was supposed to be working for the government, so imagine your error bars appropriately.
So what does this mean?
Well, let's make a couple assumptions for conversational purposes. Looks like people were voting at about 50 votes per minute during the slow times, and up over 100 votes per minute during the fast times.
At the beginning of the 50/50 period, we would have needed 248 votes one way to see the percentage change. Votes per minute at this point were flying, at about 100 votes/minute. That's only (the equivalent of) 3 minutes of voting time that would have had to go SOLID for either Vriska or Harry to produce a 1% change we could see on the face of the poll. And while we certainly weren't watching it live, the density of our data for the next hour or so was pretty solid, and we didn't see a 1% change. We have screencaps, baby! 248 votes going one way or the other at this particular point is chump change at that speed, so it IS surprising that we didn't see more scatter around 49/51 for an hour or so before things settled in.
In fact, I would stretch to say it's unusual to have seen such an even curve as the percentages changed on their way to 50/50, in a case where a few minutes' worth of solid voters could still have an impact, but as this is the first time I've ever done anything like this on a fucking tumble poll, I am not putting money on that either.
If someone wants to pay me, I am more than happy to become a tumble poll expert!
At the end of the poll, votes were also flying in, on average about 80 votes per minute. At this point, we would have needed over 1500 votes one way or another to see a last-minute shift. That's (the equivalent of) 19 solid straight minutes of every single voter picking the same candidate. So at the end it is NOT surprising to not see any 49/51 fluctuation. This is why once polls settle in, they so rarely change. It's just statistics.
So, Sev. Was this poll botted? I don't fucking know. We all need to go outside. However, if you're going to make drama on it, please do it on this post so that I can see. I'm getting laid off in 2 days and I would love the distraction and entertainment!
And a big thanks to @handwrittenhello for all the work they put in to this, and personally for providing me with an opportunity to spend a day doing one of my favorite things: working with data in Excel!
85 notes · View notes
danielcalmdown · 3 months
Text
Tumblr media
early morning, on the way to Martinaise
6K notes · View notes
sysig · 4 months
Photo
Tumblr media Tumblr media Tumblr media
Permission to headbutt: Granted (Patreon)
#My art#UT#Sans#Papyrus#Ft. something smol and I do on a regular basis ♪#This could be Handplates or it could be classic Undertale I leave that up to you lol#I definitely picked up a lot of the style quirks lol - but there are some of the ones that I like myself! Like Papyrus' darkmode clothes lol#And Sans' shorts having the stripe in the front haha - little details ♫#Realistically it probably is Handplates tho just based on where my head's at lol - I love the Handplates dynamic :D#Handplates#I talked myself into it! Pfft ♪#I found myself relating a lot to Sans especially while rereading - I want nothing more in the world than for my siblings to be happy! <3#So I gathered up a bunch of ideas of things especially me and smol do together and this was the most obviously cute one haha#Easiest to do! Tho I did still go a little extra on this lol#I'm trying to do more digital stuff ♪ It wasn't the best art day and I'm still a little nervous to jump right in :')#Not doing any sketches on paper beforehand feels weird but I guess it is thematic in a way lol#And I'm still pleased with how they turned out hehe#It really does feel nice to be drawing them again <3#And doing silly sibling things! Hehe#I dunno how clear it is since it's so ingrained into how smol and I talk to each other lol family language!#One of us will literally just announce ''bonk'' and the other will prepare for/lean in for a headbutt haha#She is a tiny bit taller than me - it's not quite /this/ extreme but she does lean down for me! S'cute <3#I like to think Papyrus would do the same hehe ♪ Let your lazy brother headbutt you! He can only reach so far!#On minimal effort anyhow hehe#It's just a fun way to be silly together ♫♪#Also yes I did show this to her and she cosigned lol - ''Cute'' -smol
4K notes · View notes
rhinocio · 8 months
Text
Tumblr media Tumblr media Tumblr media Tumblr media Tumblr media Tumblr media Tumblr media
want you to want me
6K notes · View notes
moth-nocturna · 6 months
Text
Tumblr media
my fav twst event & my fav trio !
6K notes · View notes
canisalbus · 7 months
Text
Tumblr media
What if I told you that RoobrickMarine went and wrote an entire novella starring my 16th century dog couple? It's very canon-adjacent, well researched and thoughtfully put together, has inspired me a ton during these past months and it's now publicly available at AO3. I highly recommend it.
✦ Separation ✦
Tumblr media
6K notes · View notes
sudokuishere · 6 months
Text
Is that... is that tims boat house??? Did tim let damian crash with him at his murder shack?
Tumblr media
Like? It's this room from his murder shack
Tumblr media
That's so cute I'm dying 😭
3K notes · View notes
pyralart · 8 months
Text
Tumblr media Tumblr media
In which Luz gets isekai-ed another way
First part >> Next Part
Remember folks, look both ways before crossing the road or this might happen to YOU! This was an advert from the Road Safety Agency (it's not).
Oops, I just dropped the angst! I can reveal that it's inspired by this original post, although I'll put my own spin on it so don't think you know what's gonna happen!
3K notes · View notes
pangur-and-grim · 10 months
Text
Tumblr media Tumblr media Tumblr media Tumblr media Tumblr media Tumblr media
the classical art pins are up for pre-order! greerstothers.shop
The Hunt of the Unicorn / Unknown artist The Garden of Death / Hugo Simberg Le génie du mal / Guillaume Geefs
5K notes · View notes
delicourse · 10 months
Text
Tumblr media Tumblr media
lesbian pride moment 😳🌸
4K notes · View notes
neytui · 3 months
Text
Tumblr media
inferno 🔥
2K notes · View notes
archivebottles · 1 year
Text
Tumblr media
HeartShaped_Object
[IMG ID: A knee-up drawing of Kris from Deltarune. They are standing with their arm raised near a heart wearing very detailed armor that has a goat on the pauldron and tripp pants, along with a cape with the transgender flag colors on it. The background features various patterns against a dark blue gradient. /END ID]
11K notes · View notes
liliumaa · 28 days
Text
Tumblr media
my entry for requiem cafe's contest to go to the 4/13 party.... this took me too long.... now i sleep
988 notes · View notes
yourlocalabomination · 3 months
Text
The darkness will spare my soul.
Tumblr media
1K notes · View notes
boycritter · 10 months
Text
Tumblr media Tumblr media Tumblr media Tumblr media
silly little comic about chronic pain
[image id: an 8 panel comic
the first panel has a conversation between two people. the one on the right is a light gray, and facing away from the viewer. they are meant to represent a generic person. they are saying, "wow, chronic pain sounds like it sucks." sucks is written in a dark pink. the person on the right has tan skin, a pink shirt, and pink hair. he responds "haha yeah."
the second panel shows the same two people, with the first person saying "how do you tolerate it? i'd go crazy!" tolerate is written in a dark pink. the second person responsed "um..." and trails off
the third panel shifts to a drawing of the second person sitting curled into himself. it is entirely grayscale, except for his pink hair, which is a little more muted. the text around him reads "i tolerate it because not doing so isn't an option". 'isn't an option' is written in dark pink.
the fourth panel shows him sitting and leaning against the left edge of the frame, with his legs stretched out. it reads "because what else am i supposed to do?" supposed is in all caps and written in dark pink.
the fifth panel shows a torso up drawing of him in the bottom right corner. the text reads "i can't hope for a day i'll wake up and be better." better is written in dark pink and all caps on the left half of the panel.
the sixth panel shows him sleeping, with a brown dog near his head. the text above him reads "all i can do is make sure i wake up." wake up is written in dark pink.
the seventh panel is all text, reading "i tolerate it because it needs to be tolerable". tolerable is written in dark pink.
the eighth panel shifts back to the conversation between the two people. the man finishes his response to the other person's question with "i don't know".
end image id]
4K notes · View notes
yellbug · 1 year
Text
last night i had a dream that i was at an outdoor fair and someone asked me to sign a petition and after i signed it they were like thanks so much! go pick out a prize! so i went into their booth and it was filled with hundreds and hundreds of tiny bottles of perfume and all the perfume had names like “i saw you eating greek yogurt at the reptile house at the zoo” and “you cut down the beautiful 100 year old oak tree in your yard” and i looked up and the perfume collection was called “guys i fucking hate”
7K notes · View notes