Making the stats more interesting

[Katya]

As I mentioned in another thread, the current "Popular Questions" stats are kind of boring, because Board 5.0 is so new. However, we Board readers have the power to change that by rating some older questions from the archives.

I therefore propose that, each day, we go back and rate the questions posted exactly one month ago, two months ago, three months ago, etc., back to one year ago. If we do this for a month, we'll have rated questions for all of the last year.

For convenience, here are the links to the questions posted on July 5, June 5, etc.

http://theboard.byu.edu/questions/archives/2010/jul/05/
http://theboard.byu.edu/questions/archives/2010/jun/05/
http://theboard.byu.edu/questions/archives/2010/may/05/
http://theboard.byu.edu/questions/archives/2010/apr/05/
http://theboard.byu.edu/questions/archives/2010/mar/05/
http://theboard.byu.edu/questions/archives/2010/feb/05/
http://theboard.byu.edu/questions/archives/2010/jan/05/
http://theboard.byu.edu/questions/archives/2009/dec/05/
http://theboard.byu.edu/questions/archives/2009/nov/05/
http://theboard.byu.edu/questions/archives/2009/oct/05/
http://theboard.byu.edu/questions/archives/2009/sep/05/
http://theboard.byu.edu/questions/archives/2009/aug/05/

Please leave a comment in this thread when you've finished so that we know how many people are rating old questions.

[NerdGirl]

Done.

[Katya]

Thanks! I've finished, too.

[Laser Jock]

Also done.

[Katya]

Awesome! Now only 5 of last week's questions are in the overall top 10. This is definite progress.

[C is for]

The Board title bar today is certainly apt.

See you all tomorrow to do it again!

[Dead Cat]

I've done it for...yesterday. I'll get today after I get some sleep. I also voted on my first page of favorites and a bunch more pretty willy nilly. So I guess I'm an over-underachiever.

[Katya]

So I guess I'm an over-underachiever.

Whatever you can do is good enough. Along those lines, I was surprised at how long it took to go over the past year. Would doing just the past six months be easier for everyone?

[Katya]

Here are today's links:

http://theboard.byu.edu/questions/archives/2010/jul/06/
http://theboard.byu.edu/questions/archives/2010/may/06/
http://theboard.byu.edu/questions/archives/2010/apr/06/
http://theboard.byu.edu/questions/archives/2010/mar/06/
http://theboard.byu.edu/questions/archives/2010/feb/06/
http://theboard.byu.edu/questions/archives/2010/jan/06/
http://theboard.byu.edu/questions/archives/2009/nov/06/
http://theboard.byu.edu/questions/archives/2009/oct/06/
http://theboard.byu.edu/questions/archives/2009/aug/06/

(The missing months are due to the 6th falling on a Sunday. Remember when the Board didn't post on Sundays?)

[NerdGirl]

Done. It does take a bit of time to go through the whole year, but it's fun so it's okay.

[C is for]

I like doing the whole year. I also think that when we're done with the month we should go on to even earlier! May as well vote for the whole Board.

[Katya]

I also think that when we're done with the month we should go on to even earlier!

Agreed. After this month, I was going to suggest taking a year to go through the whole archives, day by day. (I.e., 6 Aug 2010, 6 Aug 2009, 6 Aug 2008, 6 Aug 2007, etc.) It gets easier as you go back because fewer and fewer questions were posted every day. If you've got the time to go back farther right now, go for it!

[Yellow]

Speaking of making the stats more interesting, I'm interested in some feedback.

Currently, the ranking for most popular questions is determined by adding up the votes on each individual answer. This means that a question with a bunch of pretty good answers can be considered, even if it doesn't have any single outstanding answer. It also means that a question with a lot of answers has a greater chance at being considered "popular", but they're also penalized if they only have one awesome answer and a bunch of mediocre ones. I'm not sure if this is how we want things to be ranked; maybe we do, and maybe we don't.

There are other ways we could do it. For example, we could average the votes on all responses. This would mean that a question with one great answer and a bunch of lame answers wouldn't be ranked as highly. We could also just use the maximum vote count of any response attached to the question. That would mean that it wouldn't matter how many answers a question had, but it would also mean that if you had two answers with 11 and 10 votes respectively, you'd still only have 11 votes for the whole question.

I'm not sure what the best answer is. Any thoughts?

[Marduk]

It seems like adding the sum total of all votes for a response has the side effect of favoring those questions that have multiple responses and penalizes those with only one answer, even if it is better. It seems like we're trying to bring the cream to the top, so I think the best way to do that is to only consider the highest voted on any given question, and disregard all others. This way, the top responses rise, regardless of how many answers there are.

Averaging responses does the same on the surface, but an excellent response can be dragged down by the addition of mediocre responses.

[Indefinite Integral]

Done for today!

[NerdGirl]

I like the idea of basing it on the maximum vote count of any response.

[Katya]

Third vote for "maximum vote," for the reasons Marduk outlined.

[Katya]

Done for today!

Thanks, i^2.

[Dead Cat]

Maximum voting: fourthed, or whatever number we're on now. I did my voting for the sixth of ______'s and did more of My Favorites and some of My Questions as well. I think voting for these last two too is a Good Idea.

[Yellow]

Maximum voting is now in. Or, at least, will be within 15 minutes or so; whenever the cache on that page expires.