Title: Analysis - DT Depth 2 - Profile Distribution Post by: DdmrDdmr on November 13, 2018, 04:53:26 PM 1. Introduction.
@Piggy has kindly made available some interesting raw data, that allows us to get some insights into some global Bitcointalk profile related information. Specifically, I looked at the Trust profile information. The information provided in this analysis is related to the Default Trust layout, without customization of any sort, and set at depth 2 (if that is not your setting, the numbers presented here will vary from you profile vision, and the detailed profile trust information provided in a list further ahead on this post). The base considered is that of 2.437.064 forum profiles (profiles as of a week ago roughly). Trust has 3 components (A: B /C) A is the Trust Score. B is the number of unique users who have given the dealt with person negative feedback. C is the number of unique users who have given the dealt with person positive feedback. Edit: @suchmoon posted on another thread a link to a post by @theymos, where he explains the current Trust algorithm in detail (see: Minor trust score algorithm change (https://bitcointalk.org/index.php?topic=1066857.msg11432782#msg11432782)). After reading through, I've changed the description above for B and C. I’ve included a complete list of all <> 0 Trust profiles, along with some profile information here: https://docs.google.com/spreadsheets/d/17iRvcL149--2erpOmU0VufebRINo4BFvvoYHPVb-bBY/edit?usp=sharing 2. Trust Score (A) - Distribution Trust Score (A) has the following distribution across the complete Bitcointalk profile base: https://i.imgur.com/Iw3X2GB.png Things that stand out: a) 1.940 profiles have positive trust (0,08% of all profiles), 14.969 have negative trust or unknown (0,614%), and the remaining 2.420.155 profiles have this value set to cero (99,306%). b) Positive trust is concentrated on Legendries (423), Hero Members (323), Sr. Members (293), Full Members (254), Members (215) … and strangely enough Newbies (303). c) There are even 15 cases of Brand New with positive trust, although only three have been active during 2018: MindlessElectron (https://bitcointalk.org/index.php?action=profile;u=1136003) (a self-declared bot account), nemya (https://bitcointalk.org/index.php?action=profile;u=1169572) and avanpay (https://bitcointalk.org/index.php?action=profile;u=2023422). d) Negative or unknown Trust is concentrated on newbies (7.691), Brand New (1.937), Members (1.530), Full Members (1.515), Sr.Members (1.083), Heroes (573) and Legendries (248). e) There are 45 entries will a Null value, corresponding to profiles with the value ?? (unknown - examples: winter (https://bitcointalk.org/index.php?action=profile;u= 100039) and BlockEye (https://bitcointalk.org/index.php?action=profile;u= 553066)). Top 25: Code: trust rank userId UserName LastActive URL Bottom 25: Code: trust rank userId UserName LastActive URL Since there are many accounts that do not even login, I’ve also narrowed down the data to accounts that have logged in during 2018, with a non-cero trust value: https://i.imgur.com/CFKWxUY.png a) 1.343 profiles that have logged-in during 2018 have positive trust, while 7.882 have negative trust or unknown. b) Positive trust is concentrated on Legendries (390), Hero Members (225), Sr. Members (189), Full Members (159), Members (121) … and Newbies (123). c) Negative or unknown Trust is concentrated on newbies (3.486), Brand New (675), Members (964), Full Members (1.045), Sr.Members (749), Heroes (393) and Legendries (201). d) 54,56% of non-cero Trust profiles have logged-in during 2018. I crossed the average earned merit per segment x rank, to see if there were significant differences: https://i.imgur.com/vTTI0CI.png Nevertheless, I wouldn’t really pay too much notice to the above table, since we do not know the order of events on the Negative Trust segments (i.e., many likely got their awarded merit before the negative trust, not after the negative trust). The following I think have less weight in the big picture, so I’ll just lay out the distributions without further additional comment: 3. Negative Ratings (B) - Distribution Global distribution: https://i.imgur.com/ODbSvcc.png Logged-in during 2018 distribution: https://i.imgur.com/pyVfY2k.png 4. Positive Ratings (C) - Distribution Global distribution: https://i.imgur.com/ISKIP8R.png Logged-in during 2018 distribution: https://i.imgur.com/ihVmYna.png Title: Re: Analysis - DT Depth 2 - Profile Distribution Post by: suchmoon on November 13, 2018, 05:21:56 PM Great data as usual. Just a small nitpick: [1..10], and then (10..20] would probably make more sense than [1..10) and [10..20) because of how scores "age".
E.g. if you get +1 from DT the score will gradually (over a months?) go up to 10 and stay there. So [1..10] would reflect all those with aging or fully aged +1 ratings, (10..20] would be +2 etc if there are no negatives. Edit: I'm talking about #2. For #3 and #4 you might want to use a different word instead of "score" - it's not the same score as in #2, but rather a number of positive/negative ratings, right? Title: Re: Analysis - DT Depth 2 - Profile Distribution Post by: DdmrDdmr on November 13, 2018, 05:46:22 PM <...> Thanks for the suggestions. I do have some doubt's in terms of the interpretation of each individual figure (which are now clearer to me), and I also saw that there are concentration of values on the multiples of 10 values of the score. I’ll modify the OP as per your suggestions to facilitate interpreting the information. |