Topic: Analysis – The Multilingual gene (Read 315 times)

As usual, an interesting analysis, but it seems to me that it would be interesting to expand it by introducing additional groups.
Currently, we have a very wide group of "Local and Global", which includes both users who gained only 1% merits on the local board as well as 99% merits on the local board. So someone who "accidentally" received one merit on a global group immediately falls into the same bag, with a user who 99% of his merits received on the Global board - and yet it is a completely different type of users.

I think that you could divide "Local and Global" group into subgroups like for example, "Mostly Local" (for example 66% -99% Local) and "Mostly Global" (66% -99% Global) and the group "Local and Global" at 35% - 65%.

I think that such change would make results much more accurate.

Thanks for your oversight

Yes, your're right there. I really do need to zoom in to see those tables better on my Excel (or paint them with different colours for alternate rows it the table gets large)  ... thanks for pointing it out. I've corrected it in the OP.

I was singing "We are the champions" in my head, but I think there is a mistake.

11.62 is for our Teutonic neighbours no ? I see only 7.10 for french AVGMeritLocal.

1. Introduction

Ok, I’ve made up that gene, but it represents rather well what I want to depict in this post:
 
A few weeks ago, we got our first Newbie that made it to Hero, needing Merit in the process (see ). The post also shows those that made it from Member to Hero, and a few others that were on the ranking-up pipeline, very near to reaching the Hero rank. While creating that post, I was surprised to see that many of those listed posted also (and is some cases exclusively) on the Local boards.

That discovery led me to this post, where I’m going to try to depict the share that Local Boards have in accounts, being Merits de driver used to analyse the information.

A complete lists of all merited users and their stats can be obtained here, baring all the merit history in the calculus (data as of 12/04/2019):
https://docs.google.com/spreadsheets/d/1EwdBrA9GoFgEgFhVEKgvc8BF6pTBRaw1AiPrrZM8aq0/edit?usp=sharing

Columns (non-trivial):
nMeritsLocal: Number of Merits earned on Local Boards.
nMeritsGlobal: Number of Merits earned on Global (English) Boards.
nMeritsDel: Number of Merits earned on deleted posts (or on non-retrievable posts).
nMeritsTotal: Total Number of earned Merits.
nMeritedLocalBoards: Number of distinct merited Local Boards a person has been merited on.
BestLocalBoard: Local board on which the user got most of his local board merits on.
nMeritsBestLocalBoard: Number of merits obtained in the best Local Board.
PMeritsLocal: % of merits on (all) Local Boards.
TotalGroup: Total number of earned Merits (group cluster).
LocalGroup: % of earned Merits on (all) Local Boards (group cluster), excluding deleted posts.
TypeGroup: Earned merits: Only on Local Boards, Only on Global Boards, Local and Global Boards (disregarding merits earned on deleted posts in this case).

It’s rather interesting to play around with the list and filter/sort by diverse criteria (most merited, greatest %, by best local board, etc.).

Note: Deleted/non-retrievable posts should ideally be attributed to Local Boards or Global Boards in this analysis. This is not feasible, which means that some people may have been merited on certain boards which we now do not know of, since the post was deleted or is now irretrievable.


2. Overall view
An overall view of merit distribution in relation to forum members that post in Local Boards, Global Boards or both, can be seen here:



We can see above that, out of the 360.195 earned Merits as of last Friday (12/04/2019), 65,10% go to the Global Boards, 29,78% to the Local Boards, and 5,12% to posts that have been deleted (could include other concepts here too such as posts not retrievable from the scraper).

What’s also interesting in the above data is that, out of the 27.733 merited users, 15.928 (57,43%) were merited only on Global Boards(*), 8.331 (30,04%) only on Local Boards(*), and 1.910 (6,89%) on both (an additional 1.564 (5,64%) have their merits based on deleted or non-retrievable posts).

(*) I excluded deleted merits from the group creation. Some of them do have in addition some merits tied to deleted posts.

What’s even more interesting is that that the average merits per user for those that post on both kinds of boards is way above the rest of the boards: 54,08 merits/user vs 11,69 of those that post only on Global Boards and 7,83 on Local boards exclusively.
Now all those groups have a fair share of (statistical) standard deviation, so interpretation is delicate, especially when one of the groups is significantly smaller than the others.

The above is not to be taken verbatim: Posters merited on both local board and global boards are not merited per se on average five times more than those who only post on global or local boards. The above figure is not breaking down what marginal contribution each kind of board has to the average. It does denote though that many of the better merited forum members do have something to do with local boards to some extent.


I then delimited the data to the top 100 vs rest, and the result is this:



As we can see, the top 100 merited people consist of those merited on both Local and Global boards (43%), Only Global (54%) and Only Local (3%). That distribution is fairly different from what we was on the first chart, and shows the importance of local boards in combination with global boards for the top 100 merited people.
Nevertheless, the local merits represent only an average of 13,76% for them, vs 84,49% of the global boards. Average merit/person is similar for Local and Global and Only Local groups, and lower for those Only Local (but the group is very small).


3. Number of Merited Local Boards per person



The above image shows how many Local Boards a person has been merited on. As expected, those merited on Local Boards are almost always merited in a single Local Board. There are nevertheless a few exceptions. Specifically, 74 people have been merited on two Local Boards, 4 on three Local Boards, 2 on four Local Boards and 1 on, … roll of drums …,  9 Local boards.

All the data can be seen in the file linked in the introduction (simply order by nMeritedLocalBoards descending). The profiles merited in more than two Local Boards are:
@ Leteravian has been merited on the following Local Boards: Arabic, Croatian, Greek, Italian, Japanese, Polish, Portuguese, Spanish and Turkish, and has been inactive for a month now.


4. Percentage of Merits earned on Local Boards


The above shows the number of people that have a certain range of earned merits (rows), and what percentage of those merits come from local boards (columns). I’ve excluded from the calculation the merits earned on deleted posts, so a person with merits on both local board and deleted will show as 100% local on this table.

For example, for those 76 people who have earned over 500 merits, 40 have earned no local board merits (and are therefore Only Global), 13 have less or equal to 10% earned on local boards, and only 1 has earned all his merits on Local Boards (@mole0815).  

Roughly, for those merited on Local Boards, it’s a rather even 50:50 between those that have earned there under half of their merits, and those that have earned over half of their merits there.


5. Average Merits per Best Local board (per person)



The above image shows the best Local Board distribution on merits, and the average per person merited on them. I’ve only considered here the best local board per person (so the total is a few hundred merits short).

For those merit exclusively on Local Boards(left side table), the highest averages go to the French German (11,62), Turkish (11,26), and Italian (10,91) boards. The lowest are Hebrew (1), Indian (2,18), Korean (2,23), and Romanian (2,81).

For those merit both on Local Boards and Global Boards, the highest averages go to the Romanian (111,47), Dutch (101,35) and Japanese (98,33)  boards. That is on average around six times the figures of those that are only merited on Local Boards.

Now we have to be very careful here, since those averages are influenced by many amongst those top merited. For example, @LoyceV has been merited on the Dutch Local board (1 merit) and therefore drives the average pretty high there (since he is in the Local and Global group). @Suchmoon has received 5 merits on the Russian Local Board, but has less influence in the Russian Average.


6. Final consideration

We could take the top n forum members out of this analysis, but then we lose precisely many heavyweights that got merited on both Local Boards and Global Boards, and therefore cripple the "multilingual gene", which is what I wanted to see here. The message is that we must not take those averages verbatim, but people that are merited on both local boards and global boards have, in general, better averages than those that are only global board or only local board.

It’s also interesting to compare the average merits on the left hand side table, with the AvgMeritsLocal or the right hand side table, and AvgMeritsLocal vs AvgMeritsGlobal on the right hand table.