Hi, I'm from MetaFilter and I could overthink a plate of 10 - 20 beans!

June 13, 2011 1:56 PM

I noticed that the posting guidelines suggest looking at posts with 10 - 20 comments as examples of posts that spark a decent amount of discussion. These numbers seem out of date to me. Maybe they should be changed to 40 - 50 comments?

These days we seem to have at least a couple posts a day that top 100 comments, and posts that only receive 10 - 20 I would say are actually less indicative of what to do to get a lot of discussion going. What constitutes a good amount of comments anyway?

Please note that I don't believe that more comments necessarily equals a better post, or that a post with a low number of comments is a failure. Just noticed this and figured I'd throw it out there.

posted by yellowbinder to Etiquette/Policy at 1:56 PM (37 comments total)

What constitutes a good amount of comments anyway?

42?
posted by desjardins at 2:11 PM on June 13, 2011 [3 favorites]

Not a bad idea to revisit some of our templates once in a while. Thanks for the nudge.
posted by jessamyn (staff) at 2:14 PM on June 13, 2011

Grumble mumble.

Looking to start a conversation leads to a great many of the absolute worst MeFi posts, and very few of the best.

Looking to share something awesome that you found on the web leads to some great discussions, some nearly empty comment threads, and lots of great FPPs. Unless something is so stunningly awesome that it leaves us speechless, there'll be a discussion on your FPP if it's good.
posted by klangklangston at 2:18 PM on June 13, 2011 [7 favorites]

It's not quantity, it's quality. Just like whisky and dicks.
posted by Decani at 2:28 PM on June 13, 2011 [2 favorites]

I have never had quality whisky dick.
posted by klangklangston at 2:32 PM on June 13, 2011 [3 favorites]

The infodump oracle (and a little R-fu) tells me the following:

In the past year, the...
...mean number of comments on a post was 54.
...median number of comments on a post was 34.5.
...modal number of comments on a post was 8.
... and some selected percentiles are:
10% 9.0
20% 15.0
30% 20.0
40% 27.0
50% 34.5
60% 44.0
70% 56.0
80% 75.0
90% 114.0

Given that 87% of posts now have more than 10 comments, and 70% have more than 20, it seems like those are no longer very good numbers.
posted by Philosopher Dirtbike at 2:35 PM on June 13, 2011 [1 favorite]

The number should be a dynamic floating average of double the average number of comments in deleted threads (with a threshold of 10).
posted by blue_beetle at 2:37 PM on June 13, 2011

Holy shit these posts have comments?
posted by shakespeherian at 2:42 PM on June 13, 2011 [2 favorites]

I have never had quality whisky dick.

Gotta use a good dry rub, let it sit overnight.
posted by Brandon Blatcher at 2:45 PM on June 13, 2011 [3 favorites]

I always thought the indication of a good post was one that garnered more favorites than comments. I say this because those are really fucking hard to pull off.

At least for me.
posted by quin at 2:51 PM on June 13, 2011

It's easy. Post a double that gets deleted immediately, and then favorite it.
posted by shakespeherian at 2:52 PM on June 13, 2011

what really is a decent amount
of sea salt for the rub.
posted by clavdivs at 2:56 PM on June 13, 2011

In the past year, the...
...mean number of favorites on a post was 19.
...median number of favorites on a post was 13.
...modal number of favorites on a post was 6.
...selected percentiles:
10% 4
20% 6
30% 8
40% 10
50% 13
60% 16
70% 20
80% 26
90% 38

Interestingly, the favorites a comment receives and the comments it inspires are only mildly related; the correlation between log(favorites+1) and log(comments+1) is only 0.23.
posted by Philosopher Dirtbike at 2:57 PM on June 13, 2011

There are many posts that I enjoy without having anything I deem interesting to say about them.

This metric is false.
posted by Trurl at 3:11 PM on June 13, 2011

I should add: I'm glad the question came up of how we can tell if a post of ours was worth making.

6+ favorites and/or 40+ comments have been my general benchmark- but maybe I set my sights too low.
posted by Trurl at 3:14 PM on June 13, 2011

Yeah, 40-50 is probably a better range to give on that, and we should mention favorites at least a little bit (or we could just say look at the /popular page)
posted by mathowie (staff) at 3:26 PM on June 13, 2011

And here I was just trying to make enough comments to feel good about posting an FPP. Sigh.
posted by SMPA at 3:40 PM on June 13, 2011

Thinking about this, I was playing with the idea that a good post is one which inspires comments, which are then favorited. To me, this says there is a lot of interaction and high-quality posting going on in the thread. I looked at the posts which have the highest number of median favorites for the comments in the past year. Here they are (they all have a median of 3 favorites per comment):
94274, comments: 38, favorites: 32
95943, comments: 377, favorites: 54
97761, comments: 51, favorites: 60
98215, comments: 73, favorites: 27
99794, comments: 106, favorites: 24
99803, comments: 61, favorites: 20
100750, comments: 180, favorites: 41

What they all seem to have in common is that they all deal with something important to mefites (class, religion, gender). The original posts may not have been the best, ever, but they were good enough to spark active conversation.
posted by Philosopher Dirtbike at 3:58 PM on June 13, 2011 [4 favorites]

a good post is one which inspires comments, which are then favorited

Many of the best posts meet this condition.

But it's not predictive. You'll also see this in 2 Minute Hate threads.
posted by Trurl at 4:06 PM on June 13, 2011

quin: I always thought the indication of a good post was one that garnered more favorites than comments. I say this because those are really fucking hard to pull off.

That's simple: provide buckets of information and/or links. If you cover everything, there's nothing left to say. And that makes for a museum, not an active forum of discussion.

I don't think you should strive to make posts with many favorites and few comments, but rather make posts that you find interesting, with the hopes that other people will share interesting insights, comments, and information.
posted by filthy light thief at 4:15 PM on June 13, 2011 [2 favorites]

What is best in Metafilter? A mix of many things. Too many of one thing, and it becomes a niche site. Metafilter is a niche for people who like to discuss many topics, and sometimes read interesting articles.

And cats.
posted by filthy light thief at 4:18 PM on June 13, 2011

I always thought the indication of a good post was one that garnered more favorites than comments.

We discussed this on MeTa a while ago (my only MeTa post till date). cortex said it should be called "referenceability", but I preferred "favnocom" (suggested by unliteral). That thread has a bit of FishBike's fantastic datawankery, if you are interested.
posted by vidur at 4:32 PM on June 13, 2011

Lots of comments don't necessarily mean the best conversation.
posted by crunchland at 5:15 PM on June 13, 2011 [1 favorite]

Lots of favorites also don't necessarily mean the best conversation.
posted by jessamyn (staff) at 5:42 PM on June 13, 2011

According to FishBike's statistics, it seems the best way get a high number of comments and favorites is to post something about politicians, religion, atheism and rape. If you can work in something about vegetarian recipes, that would help with favorites.

So the ideal MeFi post would be something like "Prominent politicians, religious leaders and atheists discuss rape and vegetarianism (bonus vegetarian and non-vegetarian recipes) [more inside]." It should also generate at least one MeTa. And MeTas are always a sign of post quality.

(If you want two or MeTa threads, you can try something like "Prominent politicians, religious leaders and atheists discuss rape, iPods and hunting.")
posted by nangar at 6:29 PM on June 13, 2011

Thinking about this, I was playing with the idea that a good post is one which inspires comments, which are then favorited. To me, this says there is a lot of interaction and high-quality posting going on in the thread.

To me, it says it's outrage of the week topic and the typical snarky comments got all the love.
posted by cj_ at 6:35 PM on June 13, 2011

I have never had quality whisky dick.

If you insist on hanging around Scots, nothing's gonna change.
posted by yerfatma at 7:15 PM on June 13, 2011

With this sort of inflation, I will never be able to cash these favorites in for the Def Leppard mirror I've had my eye on since I started coming here.
posted by not_on_display at 9:25 PM on June 13, 2011 [1 favorite]

There are many posts that I enjoy without having anything I deem interesting to say about them.

Do you favourite them, at least? As someone who's a reluctant poster on the Blue, I find myself really needing something tangible -- either comments or favourites -- to convince me that anyone's connecting with or appreciates my posts. Is this unusual somehow?
posted by thisjax at 1:09 AM on June 14, 2011 [1 favorite]

I have never had quality whisky dick.

Opted for quantity instead?
posted by dubold at 2:25 AM on June 14, 2011 [1 favorite]

jessamyn: "Lots of favorites also don't necessarily mean the best conversation."

Ain't that the truth.
posted by bwg at 3:08 AM on June 14, 2011

Lots of anything doesn't necessarily imply the best conversation, but that doesn't mean it isn't correlated with it. That's what we're looking for, right? A way to give prospective posters something to look at before they post, to give them a better idea of what constitutes a good post? In that case, if one says "Posts with X are generally better than posts without X; look at those," and ask them (as always) to use their judgment, we've helped them.

Even a rough guide will be better than only "you know it when you see it," which is what it would be without a guideline.
posted by Philosopher Dirtbike at 3:34 AM on June 14, 2011

Well, you've run up against a basic problem with the data that you've got — while favorites could be correlated with quality of conversation, we don't know how strong that correlation is or what qualifies as a good conversation. Likewise, number of comments.

If you wanted to find out, you'd likely have to survey the userbase and ask members, say, what their top ten conversations were on MeFi. There'd be some sampling bias, because some folks would inevitably find and remember those conversations that they'd favorited heavily. You could cut some of that out by priming that question with previous prompts about what makes for a good conversation — "reasonable exchange of views," "I learned something," etc.

Then you'd have a set of n conversations that MeFites thought were good, and from there, you could see how strongly that correlates with number of favorites or comments or the distribution of comments or favorites.

But right now, the question is too inherently subjective to get a broad measure of the community. Maybe if Matt starts giving grants or starts the Journal of Metafilter Research, we can have more actual scientists interested.
posted by klangklangston at 9:21 AM on June 14, 2011 [1 favorite]

I have never had quality whisky dick.

Is that one of those British desserts?
posted by maryr at 10:20 AM on June 14, 2011 [1 favorite]

On a less stupid note, I'd say 40-50 is a sweet spot because there is/was an active conversation and it's still not too overwhelming to read all of the comments.
posted by maryr at 10:21 AM on June 14, 2011

Thirty-ninth.
posted by herbplarfegan at 10:42 AM on June 14, 2011

According to FishBike's statistics, it seems the best way get a high number of comments and favorites is to post something about politicians, religion, atheism and rape.

So basically Geert Wilders, then.
posted by topynate at 2:19 PM on June 14, 2011