Comments on: How to Build a Popularity Algorithm You can be Proud of

By: Quora

Quora — Wed, 07 Dec 2011 04:07:50 +0000

How was the StackOverflow reputation system designed?…

I can’t speak to their algorithm specifically, but I’ve been doing some work creating a similar algorithm for a web-based project. Basically, you want to count the number of “votes” that a post gets, then divide it by the amount of time that the po…

By: linkibol

linkibol — Sun, 21 Aug 2011 15:18:07 +0000

Hi guys,

I have been truly busy and was not looking at this blog… ahem… “for a while”.

Thank you for your wonderful comments.

@tripzilch:
as far as my observations are concerned, wilson’s formula is a little computationally expensive than baye’s. But it all depends on the dataset and the change in that data set (amount of new votes per day, amount of new links added… etc).

@Annabel
I plan to write as soon as I have time.

Cheers.

By: Annabel

Annabel — Sat, 30 Jul 2011 06:38:52 +0000

fantastic post, very informative. I wonder why the other experts of this sector don’t notice this. You should continue your writing. I am sure, you have a great readers’ base already!

By: Courtney

Courtney — Sat, 28 May 2011 14:57:01 +0000

This is a very detailed and helpful article. Not many will justify things in as much detail as you have, thank you

By: teoseattle

teoseattle — Thu, 26 May 2011 08:19:24 +0000

Very nice and useful post! I was wondering exactly of this stuff works and you explained me in such clear fashion. Thank you!

By: Tobin Harris

Tobin Harris — Thu, 13 Jan 2011 20:55:42 +0000

Awesome article.

Any ideas what approach to take for a series of debates, where you want to know which debate is most actively discussed?

Imagine there can be 100′s of debates a day. And each can have 0+ messages posted against it.

Basically, we want to tell users what is being actively discussed – what’s popular right now.

We also want to do ageing, because if a debate falls quiet then it’s obviously not currently popular.

By: tripzilch

tripzilch — Thu, 23 Dec 2010 10:22:41 +0000

Really cool explanation! I had already read How Not To Sort By Average Rating, which talks about the Wilson formula. What are its downsides that made you pick the Bayes formula over the Wilson formula? (the latter, to me, seems a littlebit more statistically sound). Of course you tweaked the Bayes formula to account for age and karma, but that's possible for the Wilson formula as well. Plus, if you do it right, the tweaks would be statistically meaningful as well (which is good for robustness, among other things).

By: Sebastien Ros

Sebastien Ros — Thu, 04 Nov 2010 21:44:25 +0000

This reminds the “exact” same issues with portfolio management performance calculation, nicely done.

By: Nikhil

Nikhil — Thu, 03 Jun 2010 11:48:50 +0000

An enlightening article indeed.
Somehow I am getting a hunch though that a better “damping” process can be achieved through using ideas similar to ones used in “ant colony optimisztion” algorithms.

By: MathGeek

MathGeek — Fri, 28 May 2010 04:28:26 +0000

Is this what Google uses?