Author Topic: Pareto-optimal NSF posters (Read 9193 times)

LouScheffer · « **on:** 01/29/2018 08:09 pm »

Looking at posts, it's very clear there is a tradeoff between post quality and post quantity. How you trade these off is personal, but one take is computing the envelope of Pareto-optimal posters. Post quantity is well defined, and we can estimate post quality as the ratio likes/post. Then a Pareto-optimal point is when there is no-one to the upper right - everyone who posts more has lower quality, and everyone with higher quality posts less. Making a guess from the forums I follow, I plugged in a bunch of members and got this curve. As far as I know, there is no-one to the right of the curve connecting these points, but there could be good, prolific posters in forum subsections I don't follow. I'd be happy to plot any suggestions, or you can compute the locations yourself from the formulas on the axes.

EDIT: There is some confusion over the meaning of this plot. The meaning is NOT that the lower right is bad - precisely the opposite. These are the very BEST people, for every posting frequency. For example, out of all the people who post as often as Jim does, Jim has the highest likes/post. The 'worse' posters are everyone else on the forum, who are all to the left of the curve connecting these points.

RonM · « **Reply #1 on:** 01/29/2018 08:47 pm »

You have defined post quality as likes per posts. Since these values are never negative, how do you get a negative result? Lowest possible result would be zero.

What about the x-axis scale?

Why didn't you use total likes and total posts you can easily find?

Ictogan · « **Reply #2 on:** 01/29/2018 08:54 pm »

Quote from: RonM on 01/29/2018 08:47 pm

You have defined post quality as likes per posts. Since these values are never negative, how do you get a negative result? Lowest possible result would be zero.

He took the logarithm of the likes per post, which can very well be negative.

Rocket Science · « **Reply #3 on:** 01/29/2018 09:02 pm »

Problem is that some members have been here a long time before "likes" came in to being... If you're a SpaceX fan-boi and continue "talk them up" your likes will "red-line"... I know of one member who reinvented his profile when the likes came online so that he now basically "gets likes with every post" as if he is some kind of oracle by other members commenting on that very fact...

As for me, I give them out as candy for encouraging posting without fear of expressing oneself and a way of saying thank you...

RonM · « **Reply #4 on:** 01/29/2018 09:56 pm »

Quote from: Ictogan on 01/29/2018 08:54 pm

Quote from: RonM on 01/29/2018 08:47 pm
You have defined post quality as likes per posts. Since these values are never negative, how do you get a negative result? Lowest possible result would be zero.
He took the logarithm of the likes per post, which can very well be negative.

Ok, now I see it.

Quote from: Rocket Science on 01/29/2018 09:02 pm

Problem is that some members have been here a long time before "likes" came in to being... If you're a SpaceX fan-boi and continue "talk them up" your likes will "red-line"... I know of one member who reinvented his profile when the likes came online so that he now basically "gets likes with every post" as if he is some kind of oracle by other members commenting on that very fact... As for me, I give them out as candy for encouraging posting without fear of expressing oneself and a way of saying thank you...

Yes, this is biased towards new members. Since people have different reasons for giving likes, it's not necessarily an indication of quality.

Michael Baylor · « **Reply #5 on:** 01/29/2018 10:02 pm »

Interesting idea, but you need a much more diverse group of posters if you want to prove any connections between the number of posts and the quality. F9man and SpaceX_MS obviously have very good quality ratios because they are insiders who occasionally help provide guidance in L2. On the other hand, Chris is managing the forums so he has to do a lot of the moderation work. Not really a fare comparison. Chris has plenty of high quality posts.

Rocket Science · « **Reply #6 on:** 01/29/2018 10:50 pm »

Quote from: RonM on 01/29/2018 09:56 pm

Quote from: Ictogan on 01/29/2018 08:54 pm
Quote from: RonM on 01/29/2018 08:47 pm
You have defined post quality as likes per posts. Since these values are never negative, how do you get a negative result? Lowest possible result would be zero.
He took the logarithm of the likes per post, which can very well be negative.

Ok, now I see it.

Quote from: Rocket Science on 01/29/2018 09:02 pm
Problem is that some members have been here a long time before "likes" came in to being... If you're a SpaceX fan-boi and continue "talk them up" your likes will "red-line"... I know of one member who reinvented his profile when the likes came online so that he now basically "gets likes with every post" as if he is some kind of oracle by other members commenting on that very fact... As for me, I give them out as candy for encouraging posting without fear of expressing oneself and a way of saying thank you...

Yes, this is biased towards new members. Since people have different reasons for giving likes, it's not necessarily an indication of quality.

The ones I don't get are the members who have given "1 like" in total ever... What, they gave one to themselves and that's it...

So perhaps not so much a mathematical analysis is in order but a psychological one...

Jim · « **Reply #7 on:** 01/29/2018 11:20 pm »

Quote from: LouScheffer on 01/29/2018 08:09 pm

Looking at posts, it's very clear there is a tradeoff between post quality and post quantity.

Nonsense

Lar · « **Reply #8 on:** 01/30/2018 01:17 am »

Quote from: Jim on 01/29/2018 11:20 pm

Quote from: LouScheffer on 01/29/2018 08:09 pm
Looking at posts, it's very clear there is a tradeoff between post quality and post quantity.

Nonsense

I liked that post, because reasons.
Also, as a note, you can't like your own posts.
Also, as another note, it's possible to get your post count reset.

I want to see a third variable.. likes GIVEN.

Jim · « **Reply #9 on:** 01/30/2018 01:23 am »

Quote from: Lar on 01/30/2018 01:17 am

Quote from: Jim on 01/29/2018 11:20 pm
Quote from: LouScheffer on 01/29/2018 08:09 pm
Looking at posts, it's very clear there is a tradeoff between post quality and post quantity.

Nonsense

I liked that post, because reasons.
Also, as a note, you can't like your own posts.
Also, as another note, it's possible to get your post count reset.

I want to see a third variable.. likes GIVEN.

I have had my post count reset twice, it is actually, 20k-25k higher.

I don't use the likes, I have only given 313

LouScheffer · « **Reply #10 on:** 01/30/2018 02:07 am »

There is some confusion over the meaning of this plot. The meaning is NOT that the lower right is bad - precisely the opposite. These are the very BEST people, for every posting frequency. For example, out of all the people who post as often as Jim does, Jim has the highest likes/post. The 'worse' posters are EVERYONE ELSE on the forum, who are all to the left of the curve connecting these points.

Chris Bergin · « **Reply #11 on:** 01/30/2018 02:40 am »

This thread is pointless as we've only had "likes" a year and it's no reflection on post quality. Someone could post one line of good news and it'd get 20 likes. Someone could post a mini-essay overview of an engine and it could get five likes. Jim got three likes for posting "nonsense" in his typical one word post style.

It's just a function to show appreciation.

Locked for pointlessness.