LinuxQuestions.org

LinuxQuestions.org (/questions/)
-   LQ Suggestions & Feedback (http://www.linuxquestions.org/questions/lq-suggestions-and-feedback-7/)
-   -   LQ reputation as a percentage (http://www.linuxquestions.org/questions/lq-suggestions-and-feedback-7/lq-reputation-as-a-percentage-829377/)

Berhanie 08-30-2010 03:17 PM

LQ reputation as a percentage
 
This regards the LQ reputation system, still being debated here. At least two people have suggested that using a percentage-based rather than a point-based scoring system for reputation (see here and here). The idea being that a percentage score gives an absolute measure of a person's reputation, similar to the way in which a seller's reputation is used on ebay: a person with a 100% reputation is one that can be trusted absolutely and one with a 0% is utterly untrustworthy. This may be more meaningful than a score based on points.

This post is meant to stimulate further discussion, but my submission is that reputation be calculated as follows:
Code:

reputation = ( num_helpful_posts/(num_helpful_posts + num_unhelpful_posts) ) * 100
To make this formula effective in its purpose of measuring reputation, it would require a change in the LQ culture, namely, that readers of posts would need to become more active in leaving feedback, either positive or negative.

Comments?

jeremy 08-30-2010 03:21 PM

To reiterate what I said in the linked thread, I would see this more as an improvement to the current helpful system as opposed to impacting the reputation system in any way... but I'm open to feedback on the topic.

--jeremy

Berhanie 08-30-2010 03:27 PM

Is there a need for two parallel systems? Intuitively speaking, there should be a lot (if not complete) overlap between the two.

MTK358 08-30-2010 03:35 PM

+1 for a percentage system.

xeleema 08-30-2010 03:38 PM

Quote:

Originally Posted by Berhanie (Post 4082743)
At least two people have suggested that using a percentage-based rather than a point-based scoring system for reputation ...The idea being that a percentage score gives an absolute measure of a person's reputation, similar to the way in which a seller's reputation is used on ebay: a person with a 100% reputation is one that can be trusted absolutely and one with a 0% is utterly untrustworthy....
Code:

reputation = ( num_helpful_posts/(num_helpful_posts + num_unhelpful_posts) ) * 100

Although the idea is an interesting one, I'm not at all for it.
Percentages calculated out of an average (or a mean) fail to show the whole picture. For example; grades (in my region) are typically doled-out as X of Y points (Y = an arbitrary number, typically 70). This usually results in a "bell curve" (unintentionally). The temptation to "adjust" the bell curve usually sets in.

A "percentage of trustworthiness" requires there to be a yard stick of some sort in the first place. This would require a cap.

Forgive me for being blunt, but to stick with the D&D analogies previously used in this topic;
Level Caps suck.

The current Reputation system, as it stands does not have a (permanent) cap. This is a Good Thing(tm).

On a side note, the formula you cited (understandibly a raw outline) would be *very* easy to "Game". Attempts to correct any exploits would typically be to adjust the "Cap". Shifting everyone's "Trustworthiness" in the process.

Quote:

Originally Posted by Berhanie (Post 4082743)
To make this formula effective in its purpose of measuring reputation, it would require a change in the LQ culture, namely, that readers of posts would need to become more active in leaving feedback, either positive or negative.

Even the slightest change to a shoreline requires a massive effort.
I don't forsee that happening overnight, nor in the near future.
Quite honestly I see that as counter-intuitive to most human nature, as we don't innately desire to rate every. single. thing. we. see.
Especially if we're in a hurry.

The less we require/ask someone to learn about our community prior to participating, the easier it will be.

For Example; Imagine if the "Reply to Thread" ran a spell/grammar checker prior to submitting a post, then simply returned "You have an error" when it came across something disagreeable.

I know I would quickly grow frustrated at such a dramatic shift of "Standard Operating Procedure".

Just my $0.02 USD.

Good Luck!


EDIT: Also, "trustwothiness" would probably not be the best discription. How about "technical accuracy", "Guru-ness", or "Geekification"?
After all, I know most would not suggest someone "rm -rf /" to solve their problem...well, at least not seriously.

jeremy 08-30-2010 03:45 PM

Quote:

Originally Posted by Berhanie (Post 4082756)
Is there a need for two parallel systems? Intuitively speaking, there should be a lot (if not complete) overlap between the two.

They're more complimentary then parallel. They serve two different functions and are useful independently, but I think in conjunction we can get more out of them together than the sum of the parts. I can say that replacing the reputation system with what you've outlined isn't going to happen. Within the framework of the helpful system however, having some kind of helpful quotient could be useful (and when viewed in addition to a members reputation has some interesting potential I think).

--jeremy

slakmagik 08-30-2010 04:40 PM

Hm. Part of what I said regarding the help/rep systems on the original thread should have probably been moved here.

jeremy - you may have done so (perhaps numerous times :) ) already, but could you say what the different functions are? It seems to me a persons rep should be built on helpful posts and helpful posts should lead to positive rep. So I'm not sure I'm seeing the distinction or need for the dual (or complimentary) systems.

MTK358 08-30-2010 07:16 PM

1 Attachment(s)
I'm not sure about the idea of helpful vs. unhelpful, I think that helpful vs. all would be better because I save voting someone as unhelpful for when they said something bad, not when it was just neutral.

Maybe something like this:

jeremy 09-01-2010 12:24 PM

Using "all" is going to end up being punitive to the vast majority of LQ members who have many posts before the system was implemented.

Just to get an idea what a "Helpful Quotient" would look like for some existing members, I've randomly selected a few that have participated in the ongoing rep thread, and here is what their HQ would like like (with the denominator in parenthesis):

jeremy - 87 (46)
Berhanie - 100 (1)
xeleema - 95 (20)
slakmagik - 100 (6)
anishakaul - 90 (21)
MTK358 - 100 (5)
JZL240I-U - 100 (1)
XavierP - 89 (9)
damgar - 91 (11)

With some real world examples out there now, I'd be interested in some further feedback.

--jeremy

H_TeXMeX_H 09-05-2010 10:55 AM

Well, I would say that such a system seems to produce lower scores for people that have been here longer ... is that good ?

I still don't get exactly how these are supposed to be calculated, can someone provide a full example calculation and where the numbers are coming from ?

xeleema 09-05-2010 11:18 AM

(Paraphrased & emphasis added)
Quote:

Originally Posted by jeremy (Post 4085065)
...Just to get an idea what a "Helpful Quotient" would look like...(with the denominator in parenthesis):

jeremy - 87 (46)
Berhanie - 100 (1)
xeleema - 95 (20)
slakmagik - 100 (6)
anishakaul - 90 (21)
MTK358 - 100 (5)
JZL240I-U - 100 (1)
XavierP - 89 (9)
damgar - 91 (11)

With some real world examples out there now, I'd be interested in some further feedback.

--jeremy

Um. No, not cool. I know I see XavierP as being way more helpful than myself. However I'm guessing that wasn't always the case and is skewed because of the longer timeline for XavierP.

I'm in love with the little green bags bars!
(/queue Reservoir Dogs theme song)


P.S: Level Caps Suck. With the current system, we don't know what the cap is, which is just fine by me. But 100 is too low a number, and doesn't reflect the diversity of our community.

Member A has 25 posts and all are helpful = 100% HP
Member B has 'Over9,000' posts and every one of them helpful = 100% HP?? That's just not cool, man, not cool!

H_TeXMeX_H 09-05-2010 12:53 PM

So, here's what I don't get:

# xeleema's Posts have been rated good 25 out of 27 times.

reputation = ( num_helpful_posts/(num_helpful_posts + num_unhelpful_posts) ) * 100

reputation = ( 25/(25 + 2) ) * 100 = 92.6
or just 25/27*100

So, why: xeleema - 95 (20) ?

I'm obviously using the wrong numbers, right ?

Anyway, I don't see a problem with a non capped system, like it currently is. I think it is more fair than a percentage system, so my vote is against a percentage system.

MTK358 09-05-2010 12:56 PM

Quote:

Originally Posted by xeleema (Post 4088689)
Member A has 25 posts and all are helpful = 100% HP
Member B has 'Over9,000' posts and every one of them helpful = 100% HP?? That's just not cool, man, not cool!

Haven't thought of it that way.

XavierP 09-05-2010 01:03 PM

Quote:

Originally Posted by xeleema (Post 4088689)
Um. No, not cool. I know I see XavierP as being way more helpful than myself. However I'm guessing that wasn't always the case and is skewed because of the longer timeline for XavierP.

To be honest, I'm a terrible person to be compared against - I've been here forever, most of my actions these days are mod actions and to cap it all, I'm not terribly fussed about my scores :) Be happy that you have a decent number and don't let the scores affect your time at LQ.

jeremy 09-05-2010 01:07 PM

Quote:

Originally Posted by H_TeXMeX_H (Post 4088751)
So, here's what I don't get:

# xeleema's Posts have been rated good 25 out of 27 times.

reputation = ( num_helpful_posts/(num_helpful_posts + num_unhelpful_posts) ) * 100

reputation = ( 25/(25 + 2) ) * 100 = 92.6
or just 25/27*100

So, why: xeleema - 95 (20) ?

I'm obviously using the wrong numbers, right ?

You're using numbers from many days after I gave my example, so will get a different end result.

--jeremy


All times are GMT -5. The time now is 09:51 AM.