🎓 Doc Freemo 🇳🇱 @freemo@qoto.org

@thsprsntdrknss@metalhead.club No, we don't

@Saederup92 No one to kill your dog for you when it barks at something in the middle of the night.

Today the prank on my roommate begins...

I wired up a bunch of tiny battery powered electronic devices that play random creepy sounds every few hours at random intervals. I built 9 of them and hid them in his room (like unscrewed wall sockets and stuck it in there, put it inside the lining of his bed, shit like that.. even put it inside the lining of his backpacks... there are 9 of them total

#prank #joke #jokes #humor #humour

@pschwede I would expect so as well.

@pschwede lol I was thinking about this all last night as i feel asleep. At first when i aw your bot it seemed simple, just count upt he score every 24 hours. But last night i realized as i made a post shortly before your bot announced that day that it really didnt work very well at all if it worked the simple way I had imagined.

I spent most of the night thinking up the most effective ways to address the issue :)

Just a reminder: dogs are better people than people.

that is all.

@pschwede No it judges a weeks worth of time int erms of score, but the only posts eligible are the ones posted 7 - 6 days ago, you repeat this every day not every week though, so every post is eligible about a week after it was made.

@pschwede well its a bit morethan that the idea beingthat since the I1 is much smaller than I2 the error will be smaller. I1 represents the maximum advantage a post can have (so in this case some posts still have a 24 hour lead). But if its calculated over a weeks time then that is just 1/7th error rather than an error of 1, so its a fraction of the error. The bigger the ratio between I1 and I2 the less error there is.

@pschwede No it isnt that, IMO I think its fine if the system is imperfect a bit. I just figure it should work well enough that the output is somewhat meaningful. Having a week interval means most of the highest scored posts would never reach the target.

So how about this as a way to solve it that is minimal effort on your part but still ok..

Pick two intervals the first being shorter than the other. You can pick whatever makes sense to you but lets pick one day (We will call this I1) and one week (we will call this I2).

At the end of every I1 cycle (every day) pull up all posts that were created based on this formula (Today - I2) to ((today - I2) + I1). Of those posts that meet that criteria tally the score for each post (this score will be accumulated over the I2 period so it is a weeks worth of likes). Then rank the top post of those. Repeat this every I1 (every day).

Lol fucked up but true

#joke #jokes #humour #humor #meme

@pschwede theinterval isnt the problem. If you did weeks or even years there would still be favoritism towards earlier posts. In fact it would make it even more significant. Compare the likes of a post thats a week old vs 5 minutes old (if calculated at the end of the week).

@pschwede I tend to get reactions throughout a whole day and most of my reactions are not in the first 15 minutes or even the first hour, its usually after a few hours since boosts take a while to gain momentum.

@pschwede You certainly dont have to. Simply my inputs and pointing out a bug. You are welcome to leave it as is or improve it, up to you.

It is a welcome addition either way.

@pschwede unless you take away most of the advantage to doing a late or early post. Which picking the 15 minute approach would do. The advantage would be too small for anyone to care.

@pschwede Well none. but which is why it isnt as good a solution as the timestamp solution.

But with 15 minutes then some posts have a 15 minute advantage over others, but **only** the ones in the last 15 minutes of the day. So not only is the advantage much smaller than 24 hours bit it would have no effect at all on like 99% of the posts. So its a good enough compromise I'd say.

@pschwede What do you mean?

@pschwede Then the other solution would be to periodically (lets say every 15 minutes) probe all posts that are 23 hours and 45 minutes old to 24 hours and save that info. then at midnight use that info to announce the winners. Would increase the traffic even more but if you are unable to get timestamps its the only other way I can think of.

@pschwede traffic is fine we have some pretty beefy systems...

but using timestamps what i would do is only count the likes within the first 24 hours explicitly then only consider posts between 24 and 48 hours old. That is the only sane way to do it i can think of

@pschwede Im not sure there is an easy way. Can you pull up the datestamp on likes and reshares or just the total count? with a datestamp it is easy.

@nerthos Well it is also the specific brand of cookie can. This specific type of cookie can is for some reason **always** the one used for sewing kits :)