I'm working on a new prediction algorithm to figure out how many posts will be submitted in a day.
Based on historical data and current trends, I'd give 60/40 odds that today's (May 12) post count will be between 2089 and 2489.
The way my algorithm works is by taking the differences in post count day by day from the last 6 years, turning them into a percentage of that day's posts, then averaging 4 days worth of those percentages over 6 years. This is intended to help accommodate for seasonal shifts. Then that average is multiplied by the previous day's post count, and used to predict the coming day's post count.
I can give a +-500 range of posts with 97.6% certainty, and a +-200 range of posts with a 58.3% certainty.
This translates to a 98.4% likelihood of predicting it within +-25%, and a 62.99% likelihood of predicting it within +-10%
It should be mentioned that these percentages are based off a 127 test sample, which is pretty small. I also screwed up some numbers, and after rerunning them in hopes of making it better, my certainty dropped. Hmm. I might need to peel off the 2014 data; it's the wrench in the works.
TL:DR and in plain English: I can guess how many posts are going to be made, but I can't tell you with much certainty. I can give you a 1000 post range for (almost) certain, but so can anyone else who takes 5 minutes to look at the data. It's the fact that I can predict a ~500 post range correctly 60% of the time has me really excited. It may not sound like much, but it's a start!
It's still being refined, but I have high hopes! I don't see getting much more than 75% certainty with my more precise criteria, but that alone would amaze me.
There's no reason this couldn't be applied to the other data sets, but I started with Posts/Day, since it's a much more tangible metric.