These two adjacent threads each have 66 posts:
Does it even make sense to try calculate the odds of such a thing happening, or i it one of those coincidences that don’t statistics don’t apply to?
I don’t think meaningful stats could be calculated; odds, perhaps. It’s the number of posts, which is more or less random, up against latest post time, which is also fairly random.
ETA: IMHO, statistical interpretation has meaning or it’s useless; odds can be inherently meaningless.
Look at it this way: Go down the list of threads in each of the forums of the SDMB. There are at least 400 pairs of adjacent threads. At least 90% of those threads have less than 150 posts. So there are at least 300 pairs of adjacent threads where both threads in the pair have less than 150 posts. The chance that any two such adjacent threads have the same number of posts is at least 1 in 150 then. So what you’re doing is testing such pairs of two numbers more than 300 times. I’m not going to do the arithmetic, but it’s clear that the probability that there are two adjacent pairs of threads with the same number of posts is much greater than one-half. There’s nothing surprising in the fact that somewhere in the lists of SDMB threads there will be an adjacent pair with the same number of posts.
If you ask: What are the chances that two adjacent threads made on July 5, 2014 will both have exactly 66 posts and some simultaneous point in the day, the answer is: Miniscule.
But if you ask: What are the chances that at some point in time (not limited to any particular day), two adjacent threads have the same number of posts (not necessarily 66), then the chances are much greater, and probably actually quite high that this might happen from time to time.
If you have about thirty-five people in a room and ask: What are the chances that two people have a birthday on August 15? Then the probability is low. If you ask: What are the chances that someone else in the room has the same birthday as me, that is essentially the same question.
But if you ask: What are the chances that some two people in the room have the same birthday (without caring what day that birthday is), then the chances are quite high (over 50%).
Wait, that’s still not right, is it? The chance that any two threads out the 300 have the same number of posts would be 1/150 or whatever.
Then, you have to multiply that or something to find out the chance that the two threads with the same number of posts are adjacent, right?
No, what I said was that the chances that any two adjacent threads have the same number of posts was about 1/150. But there are something like 300 pairs of adjacent threads. So the probability that some pair of threads has the same number of posts is greater than one-half. Finding that probability is more complicated than just multiplying by 300. I’ll let someone else do the arithmetic.
Are you familiar with the birthday problem? How many people do you need in a room for the probability of two sharing a birthday is 50% or better? Only 23.
And let’s just look through the fora. Right now, I see in GQ, two threads with 26 posts adjacent:
What was the most watched live event ever?
Male Answer Syndrome.
Then there’s two 16s adjacent:
Most isolated human in history?
Printer offline error.
And there’s a couple of single digit adjacencies (two 9s, two 3s), but we’ll discount those. So that’s either two pairs if you go by >10 or 4 pairs if you don’t care.
The birthday problem relies on combinatorics - for 20 people in a class, there are 20+19+18+… possible pairs - AA, AB AC… BC, BD, BE,…, CD, CE… So many pairs that the odds of any pair having any one of 365 birthdays was very likely, less than 50%. ( formula - N(N-1)/2=20*19/2=190 combinations)
In the “adjacent threads” problem, you are limited to the two pair being beside each other. With 300 threads, there are 299 pairs between a thread and the one below. (pairing with one above is duplicate)
Of course, as one thread’s number of posts increases (more popular than the other) they will at one point be equal. So if you check more than once, you will get a different answer each time and sooner or later you may hit a post count match.
Sure, but I’m addressing the first half of Sicks Ate’s response with the “birthday problem” reference. The other issue is that post counts are not going to be randomly distributed like they would (more or less) in the birthday problem. So, it’s going to be fairly easy to find duplicated post counts of 0 or 1 or some other small number (for example, right now I see a run of three threads in MPSIMS with 9 replies, and there is almost a run of three 26s, with two adjacent 26s and one 25 right before), but quite difficult over a certain higher number (no clue which one, though).
To be more accurate, instead of just treating number of posts as a random value, you could survey the board’s stats over time and get an average distribution of post numbers, which would affect the numbers. (That is, the number of threads with 75,722 posts is going to be quite a bit smaller than those with, say, 15)
Yep. That’s how I would start to get something more than a ballpark guess. Second, the problem needs to be more clearly defined. Is it just any two identical post counts, or are we saying “in threads with more than X replies?”
Not looking at the math, but the question points to a bias in human perception. The odds of two adjacent threads both having a post count of 66 are virtually the same as one having a post count of 67 and the one below having a count of 65…but few would take any notice of the latter.
The 66 & 66 case you noticed is exactly what is meant by the word coincidence.
I am going on about this because I know several people who make a practice of noticing such things, then declaring that “There are too many coincidences to be coincidence!” and citing this as evidence of devine intervention. I counter with “What number would be coincidence, and is only the extra ones that are woo, or all of them, and if the quota is not met, is that woo also?”
The other flaw in human perception is the belief that randomness should result in an even distribution. Actual randomness is clumpy. When non-savvy gamblers encounter such clumping they perceive it as a “streak” and wrongly presume that “luck” has made the outcomes non-random for a time.