Quote from: seoincorporation on April 17, 2018, 03:46:55 PM

<...>
Have you thought about a new statistical insight into this misuse? I mean: Can you make another thread (I know you are already tired, but your work is so amazing I'm afraid you are going to have many request Tongue

), regarding how many users are scamming the merit system?
And, besides, as you said, a comparison between the ranking activity before and after the merit system?
<...>

Thanks seoincorporation. Truth is, I need a bit of a break now to recover energy..

Regarding the nº of users scamming the merit system, I guess we could agree on various methods taking place:
a)   Transfer of sMerit between main/alt accounts
b)   Swapping sMerit (on purpose)
c)   Purchasing sMerit of others

They all end up in the Forum logs and are there for everyones view. The problem is discerning proper sMerit awarding from the above. Since awarding is by nature subjective and not subject to proof of work (pun), it needs a human viewer to check suspect cases one by one.
What can be done is, apart from current manual reporting, try to device lists of suspect cases pulled up from data aggregation.

I gave it a go a month ago (see https://bitcointalk.org/index.php?topic=3167331.0). That is a tough post to read, but to cut a long story short, the idea was to try to find circular merit; that is, users that award merit and get it back in return. Now this up to a certain point is logical, since sets of users like other peoples posts and vice-versa. I have a couple of such in my merit history.

It´s when the quantity gets high in these circular transactions that we can say we have detected a suspect. Now depending on the amount of sMerit given/received and/or number of times, we can filter the list and work on it (alas) one by one, to see if cases a and b can be detected (case type c is much more difficult since I doubt it will lead to a circular swap of sMerit).

For example, the tab labelled Given Received Merit I referred to in the above referenced thread, has at the top of the list (file created a month ago or so -> https://docs.google.com/spreadsheets/d/1Su1vulMzsbkYPOsU8llr8PrQXRPvy4oM_E218klwHh8/edit#gid=968451010):
User 377987 (https://bitcointalk.org/index.php?action=profile;u=377987) had given to user 1580039 (https://bitcointalk.org/index.php?action=profile;u=1580039) 50 sMerits over in 41 Transactions, but receives none in return.
Not surprisingly, these two accounts came up today on my post as weird.
Unfortunately, someone has to grab the list and filter it, to reduce the number of suspect candidates to apply a manual check on one by one check

On the other topic, to perform a comparison between the ranking system before and not is very difficult: This threads analysis is based on current user profile (of users awarded merit). What is used is a snapshot of some of the fields of users profiles, as of yesterday.

To do a comparison, we would need a similar snapshot belonging to some time ago (which we cannot retrieve) and do something similar but only related to activity. With this, we could compare throughput today vs throughput a year ago for example (although circumstances are very different now: many more users on the one side due to recent hype, but a merit system on the other that acts as a deterrent).

Even so, we can see with todays analysis that all those that could have ranked up and didnt due to lack of Merit would have ranked up with the old system (48,78%). I think that´s the best we can do for now...