Karma MathFollow

In another thread we talked about karma a little and Kao said he wouldn't mind receiving suggestions for karma.

n^-1Σ [K_n(n-1)(1+A)] / [(n-1)(1+A)]

Where K_n is the karma rating for that post, n is nth post, and A is a volatility variable. Each K is a simple average of the karma ratings for each post.

So what does all that math stuff mean? It means even Jophiel could potentially be rated down to subdefault in a reasonable period should we find him worthy. We all now that as it is, the current karma system makes high post count posters virtually untouchable. It also has the additional problem of having a strong "memory" in that the order of rate ups matters significantly. Someone who receives 5000 rate downs and then 5000 rate ups will have a vastly lower score than someone who receives 5000 rate ups and then 5000 rate downs.

The way this new formula works is that every post matters a constant percentage of all over posting you have previously done. For example say we set that percent to 10% (a fairly high number). Your next post would have as much karma weight as 10% of all the previous karma you had ever received. Every post you make matters X% as much as all the other posting you have ever done.

By doing this two key changes are made. First, high post count posters are no longer immovable and second it's possible to make guru even if you had a rocky start or for 100 post gurus to fall back to scholar if they are rated as unworthy by the forum.

Thoughts?

Edited, Jul 19th 2010 11:27pm by Allegory

I know that, historically speaking, this is hard to fathom coming from my fingers, but who the hell cares about karma this much, really?

Edited, Jul 19th 2010 11:25pm by BrownDuck

get laid.

I approve of this.

It is a different departure from the current system, thats for sure. The existing system is designed to be self reinforcing over time, and specifically to make it hard for people to get Guru. There should, under ideal karma theory, be only a handful of gurus, a few sages, many scholars, and maby "decent" people starting out. It is supposed to be trivially easy to get to scholar unless you annoy a fair number of people. The theory being that high post count individuals have "earned" their karma score over time, and it should get continually more difficult to move them.

In practice, there are some departures from the ideal. Mainly on the newbie end. 16 post instagurus, 16 post sub default wonders for daring to say something still valid, but annoying to someone else. Then there are the long term posters who hit 3-4K posts and are still sub scholar despite contributing just as much as that person who hit guru at 16 posts and spammed the crap out of the forum. We try to check and balance things a bit. The 3.00 reset switch, the occasional behind the scenes nudge when something stemmed from an early problem, but on the other hand I try not to move things too much because people get resentful, or if they find out it occured they brag about it and **** off all their peers who didn't get an adjustment.

The initial proposed function there would drastically increase volitility in the system, and decrese stability. I guess the question would really be, is that a good thing? 4 or 5 users could send an individual Jophiel post sub default right now, the next post he made would be back to excellent. Under the more volitile system, a few people could effectivly suppress the posts of 1 individual indefinitly, very quickly, never giving them a chance to redeam themselves.

I'd have to vote no on that one I think. It does address some of the key issues, but it could create some monster problems that would be hard to fix.

I've put a request in to try a 100 3.00 rating initial pad on account creation. It will only affect new accounts when it goes into effect. I figure we'll try it and see how it goes. Another, more mathmatically complex option would be apply ratings to overall karma on a logaritmic scale. Still an overall average, but the earliest rates would only affect your score a fraction of what rates would affect your score the more posts you have. (mathmatically flawed example: posts 1-10 would only recieve 1% of a rating, 10-100, 20%, 100-500 50%, 1000, 100%) etc. That might make the servers melt though.

Or maybe give everyone a "reset your karma to 3.00" button, with 1 charge. Use it, and you are done. (or maybe we let people buy additional "reset potions" to help fund the forums? I dunno.

Should greater post count affect the weight of your oppinions? I think I say yes. Spammers should be discouraged by post docks, but people who have put in the time and ink to try and shape the forum should have a bit of a say in its desteny.

Is definitly a more complex issue once you get into it. And really, the system does work as is to a certian extent. Most people who are ******** end up at the lower end of the karma range. Most people that are nice and friendly and don't eat babies end up in the upper range. And then there are the admins... (well, exceptions exist everywhere right?)

(Actually I had my 5.00 before I was an admin, but I started before everyone else too heh)

It doesn't so much make it hard to get guru as it makes it hard to change your status at all.

I ran a test in excel (because I was lazy). I created 2 hypothetical users. Bob started out as a good poster. Bob made 1,000 posts that each received two 6.00 rate ups, but then Bob got bitter after his girlfriend left him and started getting two 1.00 rate downs for the next 1,000 posts. Bob's final karma at 2,000 posts is 4.5. Timmy had a rough start on the forum. Thinking he was an internet tough guy Timmy made his first 1,000 posts each receiving two 1.00 rate downs. Later Timmy learned the error of his ways and made his next 1,000 posts receiving two 6.00 rate ups each. Timmy's final karma is 2.49.

Both posters received the exact same number of rate ups and rate downs, but because they received them at different times one of them is a guru and one of them isn't even a scholar. Not only that, but neither rating reflects their current behavior. The default guy is now a heck of a lot nicer than the guru.

That's why this karma system is a poor representation of how the community views a poster.

The biggest problem with the existing karma system is that you rate yourself twice for your present score. This creates a snowballing effect. A simple fix would be to have each post rate itself twice for your present karma score, but this self rating doesn't factor into your overall karma score. It only exists as a basis on which other people can rate (so that it still takes more rate downs to get a guru's post subD than a scholar's). Your actual karma score should just be a simple average of all the rate downs and rate ups you have ever received. This way your score isn't affected by when you received the rate ups. Using this method Timmy and Bob would have exactly the same karma score having received the same number of rate ups and downs.

I'm mostly interested in this as an academic exercise. I don't expect any significant changes to karma to go over well with the community.

Edited, Jul 20th 2010 2:18am by Allegory

It would certainly make Karma more interesting and relevant.

This. It's the most elegant solution.

The current system is past-focused, attributing much greater weight to the past than to the present or future.

Allegory's system is future-focused, attributing much greater weight to the future than to the past or present.

Both systems require more complex tabulation calculations than a suspended average.

A normalized system is both evenly weighted between past/present/future actions, very easy to program, takes up few resources, and would let me get guru faster.

I'm with Nixnot. I know I care way to much about karma. I'll admit it.

Bring back the numbers.

And Nephty.

For someone who supposedly considers karma of no use/value, you sure do spend an awful lot of time thinking about it and discussing it here.

I like this.