Does it depend on the individual forum?
Kind of related: do most users browse through each forum individually, or is browsing via the “new posts” function more common?
Not sure there is one.
For one thing, we now have board members literally from every corner of the globe.
And not everyone lives the same sort of timeframe, so we get visited at all hours of the day and night, it’s always prime time for someone around here.
There are times on the board where we are too busy, where there is so much demand on the server that stuff bounces. That’s probably not a good time to be looking for good response. Tends to be . . . 3-6pm eastern time, folks? That’s when I have the most trouble, with the peak times seeming to be 5-ish.
And the board is unavailable for a while in the middle of the night, when the server does its backup thing. It used to only be a few minutes, but of late, it has seemed to be like it’s been considerably longer. However, the board is quite large and I would imagine as we grow the board day by day there’s more and more to handle and so it takes longer.
Can’t speak for other people, but I tend to look at each forum separately. But you do what works best for you, however that might be.
your humble TubaDiva
Adminstrator
If it means anything, I have a chart of the number of posts made for each hour of the day collected from a one-week period in mid-May (this is part of a larger study that I’m writing up). Whether this data is typical of any other week, I can’t say.
One might think that the best chance of receiving the most responses for a thread is to post it during the busiest times of day, but the problems are a) that’s also when the competition is the fiercest, and your thread may not get noticed at all, and b) as Tuba mentioned, at those times the servers are so busy that people may not be able to post responses to your thread.
There is no way to tell retroactively how many views each thread receives over the course of a day, though I’d imagine the broad pattern should be somewhat correlated to the posting rates. I do know from experience that the best way for a thread to stay on the front page is for it to be posted at a low-traffic time in a low-traffic forum.
Actually that starts at 11 AM 
Ah, but you have forgotten I live smack dab in the middle of the eastern time zone.
There is no time but eastern time. All else is just stuff that happens to someone else. It’s like the U.S. dollar.
your humble TubaDiva
“How much is that in real time?”
That’s pretty interesting. Assuming that most of the users are in the US, it appears that a significant number of posters who post during the first hour or so of the workday.
And assuming that the general pattern is valid for every week.
Near as I can figure, it wouldn’t make any difference, on average, whether you post during a busy time or a slow time. Either way, it takes the same number of posts to scoot your thread to page 2, and assuming that the distribution of posters and of post:read ratio is constant in time, that’ll also mean the same number of folks reading/posting to your thread before it reaches page 2.
Now, if my assumption of uniformity of posters is invalid, then there might be an optimum time for your thread, but it would depend on the topic. If, for instance, Lord of the Rings fans have a disproportionate tendancy to be online at 7:00 PM (just an example; I have no idea if this is true), then that would be a good time to post an LotR thread, while all the fans can read it, but it would be a worse time to post any other thread, since there would be a tendancy for the LotR threads to crowd others off the front page.
On the other hand, if there’s some time when the views:posts ratio is very high, that would be a good time to post threads you want read, but not necessarily replied to. Again, I don’t know when such a time would be, if at all, but I see no reason to suppose that it would corellate with busy or slow times.
First, a couple of assumptions:
Posters only scan the front page
Posters post in threads at random at a fixed rate dependant on the time of day
Posters start threads at a random at a fixed rate dependant on the time of day
The amount of time a thread is on the front page is t[sub]front_page[/sub] = N[sub]front_page[/sub]/V[sub]new_threads[/sub] or the number of threads on the front page divided by the rate threads are created. The average number of replies you get to your post is V[sub]replies[/sub] = V[sub]new_posts[/sub]/t[sub]front_page[/sub]. Now, if V[sub]replies[/sub] >= 1, you will be perpetually on the front page, otherwise, you will fall off and V[sub]replies[/sub] = 0. Thus, your goal is to find the longest period of time in which V[sub]replies[/sub] >= 1 and post your thread at the start of this time period
Earthling needs to come up with a V[sub]replies[/sub] graph for different times of the day so we can get a definitive answer.
Hm. If that’s the case, I’m not sure we need to do a whole lot more math, then. Your two formulas can be combined into:
V[sub]R[/sub] = V[sub]NP[/sub] / (N[sub]FP[/sub] / V[sub]NT[/sub])
or, viewed more simply:
V[sub]R[/sub] = (V[sub]NP[/sub] x V[sub]NT[/sub]) / N[sub]FP[/sub]
Since N[sub]FP[/sub] is a constant, and V[sub]NP[/sub] and V[sub]NT[/sub] are positively correlated, this just means that V[sub]R[/sub] changes as a simple multiple of V[sub]NP[/sub], which is what I have graphed. In other words, we should be able to use the general shape of my graph to estimate V[sub]R[/sub] in relative terms.
There is a catch, though – viewed over a timescale of months and across all forums, new-post and new-thread creation rates have a correlation coefficient of 0.97. When we drill down to the hour and individual forums, there may very well be instances when V[sub]NP[/sub] and V[sub]NT[/sub] diverge, but I don’t have the data to determine that.
Speaking of data, Shalmanese, I want to offer you my profuse apologies. You had asked me for some raw data way back when. I haven’t forgotten about it, but now that I’ve gone back and looked at your request again, I realize that I misread it to be asking for all the data that I’ve been gathering. :smack: The stuff you wanted, I could’ve provided much earlier, but if you’re still interested, I do now have all my other stuff available:
[ul][li]Basic_Data.xls (148 KB .ZIP) Raw sample data collected for user registrations, posts made, and threads started.[/li][li]User_Analysis.xls (270 KB .ZIP) Sorts users by post count, post rate, and active lifespan.[/li][li]Posts_Threads_Forums.xls (69 KB .ZIP) The average number of posts annd threads made per day. Sorts threads by the number of posts in each.[/li][li]Subscriptions.xls (411 KB .ZIP) Tracks subscribers through the discounted sign-up period, and a tally of more recent subscribers.[/ul][/li]Sorry, again, for the misunderstanding and delay.
Wait, something isn’t right, the formula doesn’t pass a basic sanity check. Logically, if V[sub]NT[/sub] = 0, then V[sub]R[/sub] must be > 1 since if no new threads are ever created, your thread will always stay on the front page. But the formula states that V[sub]NT[/sub] = 0 → V[sub]R[/sub] = 0 which is wrong.
Earthling: Another approach to this would be to simply verify it experimentally by plotting an average length vs time started graph, do you still have all the original data to do that? Also, if you still have the original threads, you could extract out all the posting times and perform a simulated board run, it would be interesting to see just how well my initial assumptions hold. The two major things I left out was the fact that sometimes threads from the 2nd page or lower are bumped back up to the front page and that some threads are inherently more interesting than others. It would also be nice to plot an average post rate/thread lenth graph for all threads to see how quickly a thread dies.
However, I just realised that it’s most likely you don’t have this data as you only scanned the front pages instead of each individual thread. If I can find the time, it would be interesting to write a spider if the board admins permit it (I’m in Australia so I can pretty easily do analysis during off-peak hours so server load wouldn’t be a problem). Alternatively, if the admins see fit to provide an SQL dump of a random month, that would also be good.
I’m glad you said that. The formula didn’t make any sense to me, either, but then I thought, “surely he knows what he’s doing…”
Regarding your other comments, I actually collected all the threads that were active for the last 12 months and wrote about it here. The forum view pages don’t show when each thread was started, though, so I had to estimate that based on thread-start data collected in another file. You can download the result here (3.5 MB .ZIP file, expands to Excel format). With this file, you should be able to figure out how long a thread lasts* and see if it has any correlation to what time in the day it was started.
- Because of estimation errors, there are some instances where the thread start time shows to be later than the last-post time, but I think it should be close enough for purposes of figuring out trends.
I was really more interested in the shape of the curves for each forum that just a raw length so what is really needed is data on each post rather than each thread which I assume you don’t have. Oh well.
OF course, V[sub]R[/sub] = V[sub]NP[/sub] * T[sub]FP[/sub] not divides. So you get V[sub]R[/sub] = (V[sub]NT[/sub] / V[sub]NP[/sub]) * N[sub]FP[/sub] >= 1 or V[sub]NT[/sub] / V[sub]NP[/sub] >= 1/N[sub]FP[/sub].
That makes a lot more sense.
No, I don’t have that data, but I’m actually not too clear on why it’s needed. I went back to the collection of threads from last year and was able to average the number of replies and views each thread received, based on the time of day of the OP. Though the chart makes it look like there are some large hour-to-hour variations in the reply- and view-count, we must realize that the vertical scale of the chart is exaggerated, and the max/min variation has a range of about +/- 12% – so it may not matter that much what time one submits a new thread.
I was interested in data like:
80% of the posts in a GQ thread happen within the first hour wheras only 10% of GD posts do.
Most posts show an exponential decay in posting rate that can be fitted with a 87% correlation.
theres a strong peak of posting rate 24 hours after the initial post was made.
Of course, I’m just making the above figures up.
Also, its interesting to see from your graph that there’s quite a strong correlation between posts and views. How strong is this and what’s the co-efficient? Could it be that the key to having a successful thread is to just get more people to look at it than the inherent interest in what you have to say? Could you put up an XY scatter of posts vs views?
You guys bruised my brain.
I’ve found through experimentation that the best time to post is early in the day (EST), Monday through Thursday. I’ve found Fridays to be unpredictable for obvious reasons.
It makes me smile that Friday and Saturday nights are kind of dead here. I’m glad the Dopers are getting out, of course in order to know that it’s dead then makes me a loser by default. I only check briefly on weekends, unless my roommate is hogging the TV. Why in the hell am I rambling, this forum does weird things to me.
It’s OK dude, you’re among friends.
I’m glad I was able to ask a question that used all sorts of math I can’t understand. Makes me feel kind of smart.
I know Earthling’s gotten pretty far into analyzing all sorts of board-related statistic. To what end, I wonder. (Although I recall in a previous thread, possibly the original, he admitted that he had no idea)
Ah, I see. But no, I don’t have that information. It is interesting, but would be impossibly hard to gather by hand and probably kill the hamsters in the process.
Wow, I’m glad you asked. The results are very enlightening. The correlation coefficient between views and replies for all threads active in the last year is actually not that good – around 0.35 to 0.45 or so (it’s actually a bit hard to figure out, since I have more data than Excel can handle at any one time). If we restrict the data to only threads started last year, though, the correlation improves markedly, to the 0.75 - 0.8 range. This suggests that, in general, the number of views and replies to a thread are strongly linked – and I believe that’s just intuitively sensible – but the few ancient threads that are somehow still active throw things all out of whack.
Incidentally, the correlation between either views or replies to the lifespan of a thread isn’t too great, either – it’s about 0.2 or so for all threads, and only about 0.5 for threads started last year. I’m guessing this is due to long-dormant threads getting resurrected once in a while.
That’s sort of a circular question, isn’t it? If you start a thread about a popular topic, then many people will view and reply to it. And if you write about an arcane subject, then only a few would be interested. In other words, I don’t think there is any causal relationship between views and replies; rather, it’s the subject matter itself that’s important. And this is also why mods frown on misleading thread titles.
Well, I had to sample the threads down and toss away a few outliers so Excel wouldn’t choke, but here ya go. Unless I’m seeing patterns that don’t really exist, it appears that toward the extremes of the sample, the data seems to bifurcate, with one branch showing “many replies but relatively few views” and another showing “lots and lots of views but not so many replies.” FWIW, here’s a quick and dirty views-to-replies ratio histogram (using only threads active before subscription).
Well, I like data. And I (involuntarily) have time on my hands at the moment. I’m pretty much done, though, having published the big study that took over my life. Need to get on with other things now.
Are you saying you gathered all this data by hand? A simple spider would trivially be able to collect the data I need and, unless the CR is paying by the Megabyte instead of by the Bandwidth, doing this during the ebb time (around 3am EST according to your graph) would result in no disruption for anyone.
How many threads are active but not started last year? I can’t imagine it would be that many.
How are you measuring lifespan? The common method is to go from something like the 10th percentile to the 90th percentile range to get rid off all the outliers, I suspect this is what’s fouling up your results.
Not really, We can assume that the decision to look at a thread is based solely on the title and possibly the blurb with the new version of VB. The decision to post would be presumably be based on the content as well. If the correlation is strong, it means that the decision to post is based almost solely on the title alone, a weak correlation would mean that content is more important.
Hrmm… it’s interesting that nearly all threads are restricted to within a very distinct band but seems fairly evenly distributed within that band. Could you post a zoomed graph at around 200x5000? Maybe it’s just the dot’s obscuring the true data. Alternatively, could you post the xls file? I would be very tentative about the bifurcation hypothesis, it doesn’t look like there is enough data to confirm it. Of course, you get the “Is my vagina normal” threads that generate an inordinate amount of views.
