Title: User's sMerit Network size – Data to select possible Merit Sources Post by: DdmrDdmr on May 16, 2018, 04:20:04 PM Edit 19/06/2018:
Added this information to the Merit Dashboard (Link: BitcoinTalk Merit Dashboard (https://public.tableau.com/profile/ddmrddmr#!/vizhome/BitcointalkMeritDashboard/GlobalSummary)), under tab "Potential Merit Sources" so as to allow for a dynamic periodical update of the data, as well as to apply filters to it with recalculation of the ratios (by section/subsection, date period or specific user). Data on dashboard is up to friday 15/06/2018. --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- 1. Introduction We’ve seen various post related to the Merit Network, both from a user’s perpective (Our very own sMerit Network Picture (https://bitcointalk.org/index.php?topic=3395255.msg35551989#msg35551989)) and from a global perspective (sncc's Merit network analysis: merit rank distribution and satellites (https://bitcointalk.org/index.php?topic=3759121.msg37064243#msg37064243)). The question of how to derive the potential sMerit Sources has come up recurrently, so I thought about digging into the data to see how this could be determined, or at least eased. The basic principal that I’m using is that potential users should have a track of sMerit awarding in their history. I think that the best criteria is determined by the size of the personal outgoing sMerit Network. That is, the number of distinct users that a certain user has awarded with sMerit over time. The hypothesis behind is that the more users that one has merited, the better knowledge he has of the underlying principles. The amount awarded is secondary in this case, but also a factor to narrow down candidates. This can be done from: a) A Global perspective: Aggregating all the data for a given user. b) A Section/Subsection perspective: Aggregating the data for a given user in a specific section/subsection. The numbers presented are those that I found, based on the snapshot of data collected 10 days ago (and aligned with the data presented in Forum Metrics - Section/subsection sMerit breakdown (https://bitcointalk.org/index.php?topic=3531472.msg36287248).). Since I do not know who the exact 79 sMerit Sources are, I have not taken them off the analysis. Logically, the current sMerit Sources should come up in the data right at the top in most cases. In addition to determining potential candidates for Merit Sources based on the above criteria, Merit Sources could be added by including moderators to that role, since they are already scouting the forum and parsing posts daily. It depends on whether they consider they can perform this additional task along side current tasks. This idea was nicely suggested by seoincorporation in other posts. In all cases, the candidates need to be manually screened, and vetoed if necessary (tracking DT, checking merited posts, etc.). In addition, I’d like to mention that determining the potential sMerit Sources comes in second place to determining where these sMerit Sources should develop their activity. Determining where there is a lack of sMerit Sources, in my opinion, derives from knowing how much sMerit is being given in a given section/subsection, and getting a measure of how many posts go unmerited, all in a given timeframe. The deficit can then covered by placing sMerit sources at work in those areas. The idea here is to be able to detect potential sMerit Source candidates, not to determine in this exercise where exactly they need to be active since that requires other input variables like I said. 2. Global Perspective From a global perspective, this is the distribution of the size of the outgoing sMerit Network for all users with at least 1 sMerit sent over their history: https://i.imgur.com/kMCwNNO.jpg We can see that 26 people have sent sMerit to over 100 distinct users each, 4 people to between 90 and 100 users, and so on. Only 6,63% of the users have a network with 10 or more awarded users. On the other extreme, a mass of 7.335 people have sent sMerit to only one user, representing 54,61% of all total sMerit awarders. I think that this last data is a low key tone in the system (more than half of awarders only awarded one user). Obviously this is related to the amount of sMerit available, but from the big picture’s point of view it doesn’t look great. So from a global point of view, the users that are candidates to being sMerit Sources would be (again, sMerit Sources will probably be on the lists at the top): (note: top 100 sMerit network size are provided in this list. Full list in a more user friendly format can be retrieved here:Google Drive sMerit Outgoing Network 20180516 (https://docs.google.com/spreadsheets/d/1vXLD1wk5oO8Ot9VgYKaeIgfW0-ggqeapnWTFcwSfwew/edit?usp=sharing)) The idea would be to scout the users on the list in a top down manner, favouring network size, but also using the complementary provided fields as filter criteria. Provided columns are: User_from: User_id Name: User alias Rank: User rank (as of data extraction) nSentTo: Network size. Number of distinct users that user_from has sent sMerit to. sMeritSent: aggregate sMerit sent by the user. nTx: Number of sMerit Transactions. avgSent: Average sMerit sent per Transaction. stdDevSent: Standard Deviation of sent sMerits. cVarSent: Coefficient of Variation (the smaller, the more constant value of merit awarded by the user) Code: user_from name rank nSentTo sMeritSent nTx avgSent stdDevSent cVarSent NetworkGroup Code: user_from name rank nSentTo sMeritSent nTx avgSent stdDevSent cVarSent NetworkGroup Suchmoon comes at the top of the list, having the biggest outgoing sMerit network in the whole Forum (as of date of data extraction). He has an outgoing network of 403 distinct users, having awarded them 1087 sMerits over 738 Transactions. The average per Transaction is of 1,47 sMerits, although the awarding has a variance of 1,1 sMerits giving us a Coefficient of Variation of 74%. Paxmao is next in line, with an outgoing network of 363 distinct users, having awarded them 571 sMerits over 519 Transactions. The average per Transaction is of 1,1 sMerits, with a variance of 0,39 sMerits giving us a Coefficient of Variation of 35%. This means that Paxmao tends to award an amount of sMerit nearer to his average than Suchmoon. This is neither good nor bad, but the Coefficient of Variation could be meaningful in some scenarios in general. And so on.. 2. Section/Subsection Perspective A section/subsection perspective allows us to see the same output of information as above, but filtered specifically to the section/subsection being analysed (aggregates of data are only those related to the specific section/subsection being treated). The graphs themselves are not too enlightening, but the data table at the base shows us the goodies. I'll spare the visual pain of seeing local scrolling lists here, but all the data charted can be retrieved from the full list here (second tab) :Google Drive sMerit Outgoing Network 20180516 (https://docs.google.com/spreadsheets/d/1vXLD1wk5oO8Ot9VgYKaeIgfW0-ggqeapnWTFcwSfwew/edit?usp=sharing)) 2.1. Alternate Cryptocurrencies https://i.imgur.com/2WdlV6B.jpg Vapourminer has the largest outward sMerit Network, with 64 Users in his network on the Mining Altcoins board. If we needed sMerit sources on the Mining (Altcoin) section/subsection, there are 11 candidates with an outward sMerit Network greater or equal to 10 users, so finding one there is difficult due to lack of candidates. Nevertheless, perhaps candidates with a smaller network in the section/subsection, but a large overall one could be shifted to Merit here as a source. All other sections/subsections are similar except for Ann Altcoins, where there are 45 candidates with a network above 10 users. 2.2. Bitcoin https://i.imgur.com/EsMLhvw.jpg Project Development has only two native candidates, since all the remaining users have a network that is less or equal to 6 users. Bitcoin Technical Support is also rather scarce, with only 10 candidates with a network greater o equal to 10 users. Mining has one more candidate above this threshold. Development & Technical Discussion and Bitcoin Discussion are better off with easy two digit network candidates. 2.3. Economy https://i.imgur.com/yqp9yCw.jpg Trading Discussion is a bit scarce, but holds 18 candidates with a network greater or equal to 10 users. Marketplace and economics have plenty of candidates above the 10 user network threshold. 2.4. Other https://i.imgur.com/qUSIhk0.jpg Meta seems well served with candidates. Beginners & Help has just over a dozen above the 10 user network threshold. On the other hand, politics & Society, and off-topic are rather scarce with candidates, with less than a handful at most. Archival and Serious Discussion probably don’t need candidates. 2.5. Local Boards https://i.imgur.com/w0cQbJT.jpg Half the Local boards are a sore sight in terms of candidates. Moving the minimum size threshold further down gives more options, but lower than a network size of 7 or 8 is maybe dangerous terrain due to farms. Russian, Turkish, Indonesian and Turkish local boards seem to have enough candidates if needed. Some seem impossible using a sMerit Network criteria, such as the Indian board. If it’s moderated, perhaps the mod could do this job too. Title: Re: User's sMerit Network size – Data to select possible Merit Sources Post by: mdayonliner on May 16, 2018, 04:24:56 PM Another mathematical analysis... I actually love seeing these stuffs from you and sncc. There is another one I guess. Good work guys.
Title: Re: User's sMerit Network size – Data to select possible Merit Sources Post by: seoincorporation on May 21, 2018, 02:09:41 PM <...> I've suspected for a while for Paxmao to be a Merit Source, but maybe I'm wrong. Being or not wrong, he's a perfect candidate, for, as I have seen, his merit distribution is fair and thoughtful. Taking a look at the graphics, Jet Cash is the perfect candidate to be a MeritSource, for he is truly concern about giving merits and he's always searching for good posters (even now, he'd started a "project" in which the Spanish local can receive some new readers by translating some awesome posts by Gothorum and others good posters in there). Besides, he's active on boards like "Begginers and help". I haven't seen you in the analysis. Are you out of it or I just haven't searched for it properly. For I have to speak about a candidate for the Spanish board it will be you, no doubt. The analysis is great and provides a clear vision of the most actives members in the matter of merits. I haven't given as much as I liked because I don't have many, my first amount was around 20 sMerits to give (if I recall correctly), and I am now practically merits-off. I don't consider myself as a great merit source because my English is far from perfect. Thanks for that, DdmrDdmr, as usual, your contribution is great and I'm grateful to you for explaining all those issues we have discussed into your amazing graphics. I hope it to be useful to inject some smerits into the system in order to encourage the people to continue being great. Title: Re: User's sMerit Network size – Data to select possible Merit Sources Post by: vlom on May 21, 2018, 02:21:15 PM i found my name in the list:
550439 vlom Legendary 71 165 83 1.99 2.83 1.42 (70..80] and i am not Merit Source. I only have 48 sendable merit left. did not receive much. i think it is enough to check the most genereous senders to have the list of Merit Sources. https://bitcointalk.org/index.php?action=merit;stats=topsendat but because i liked what you did i sent you some merits. Title: Re: User's sMerit Network size – Data to select possible Merit Sources Post by: darklus123 on May 21, 2018, 02:49:34 PM Great data. Not to mention that most of the names youve selected are really generous in giving merits. Tho I just dont want to settle this down by how much number of merits they have given but the fairness of there decision. I cant mention names but truly they all deserve to be or maybe some of them already are a merit source
Title: Re: User's sMerit Network size – Data to select possible Merit Sources Post by: DdmrDdmr on May 21, 2018, 05:13:50 PM <…> Yes, I figure Paxmao is a Merit Source, and a very good one at that. Being only a full Member, his sMerit Airdrop would have been only 20 sMerits at most (sMeritAirdrop = (yearActivity / 378) * modifier * freeMerit, where modifier is 0,2 for Full Members and freeMerit the Rank Merit requirement).Jet Cash is a very good candidate, with an extensive network (198 at the time of data analysis) and a very evenly spread of sMerit per Tx (1,07, with 0,36 std. dev.). LoyceV I think has just been appointed as a Merit Source with very similar data to Jet Cash. I do appear in the Code List in my OP (1582324 DdmrDdmr Member 42 58 55 1.05 0.30 0.28 (40..50]). It's easier to see the data in the complete list (every forum user who has sent sMerit is listed here:Google Drive sMerit Outgoing Network 20180516 (https://docs.google.com/spreadsheets/d/1vXLD1wk5oO8Ot9VgYKaeIgfW0-ggqeapnWTFcwSfwew/edit?usp=sharing)) P.D. Your English is way above average for non-native english posters, so don’t shy around it... (although I know it often requires an overdose of time to express the same as in the native language.) <…> I thought so too initially, but then I thought network size could/should be favoured over total sMerit sent, since it gives us a view of how well the person interacts with others on the whole in terms of sMerit awarding. In certain cases, users that award a lot but to few people may not be as good merit sources as those that award less but to many people.i think it is enough to check the most genereous senders to have the list of Merit Sources. <…> Also, if we take into consideration the forum/subforum structures, perhaps a more local approach of our sMerit awarding behaviour should be considered. For example, on the above referenced look to the complete file, the second tab has this data in your case: Code: Alternate cryptocurrencies Altcoin Discussion 550439 vlom Legendary 6 6 6 1.00 0.00 0.00 6 This is complementary to the global aggregate view, but could be a primary means of detecting potential sources on a more focalized specific section/subsection. <...> Cheers. The short code list is just an example of top candidates from a global perspective. Half of them are probably already sources, so I included the complete list of all users both from a global perspective and a more select Forum Section/Subsection.For example, using the complete file, If I were trying to scout for a Merit Source specifically for the Spanish Forum, the top 5 candidates based on filtering the file’s second Tab would be in this order: Paxmao (network size on Spanish Local of 101), Gothorum (27), solosequenosenada(15), EcuaMobi (15), and foroplus (foroplus) -> I’m personally further down the rank in this section/subsection.. The idea would be to work the list until a suitable new Merit Source is found. Further due diligence can be performed checking on the cherry-picked candidates their global network stats on the first tab of their file, and looking through their awarding and perhaps post history. Title: Re: User's sMerit Network size – Data to select possible Merit Sources Post by: LoyceV on May 21, 2018, 05:42:48 PM Jet Cash is a very good candidate, with an extensive network (198 at the time of data analysis) and a very evenly spread of sMerit per Tx (1,07, with 0,36 std. dev.). LoyceV I think has just been appointed as a Merit Source with very similar data to Jet Cash. I only saw your thread today, but I closed my Merit source application thread (https://bitcointalk.org/index.php?topic=2851620.0) just after you opened it. Coincidence or not, thanks for the endorsement.I only have one comment to this thread: if you use the sMerit Network size as a measure to determine who would make a good Merit source, your system strongly favours high ranked users who started out with a lot of free sMerit. Someone who started with with 200 sMerit can much easier merit 100+ different users than someone who started with 50 sMerit. That makes sMeritSent divided by nSentTo an indicator for lower ranked users. A lower number would be better in this context. Title: Re: User's sMerit Network size – Data to select possible Merit Sources Post by: vlom on May 21, 2018, 06:17:50 PM <…> i think it is enough to check the most genereous senders to have the list of Merit Sources. <…> I thought so too initially, but then I thought network size could/should be favoured over total sMerit sent, since it gives us a view of how well the person interacts with others on the whole in terms of sMerit awarding. In certain cases, users that award a lot but to few people may not be as good merit sources as those that award less but to many people. you are right. and it solves the problem that people are in the list just because they send a lot of merit to accounts that they are farming. done like you did it excludes them from the list. Title: Re: User's sMerit Network size – Data to select possible Merit Sources Post by: DdmrDdmr on May 21, 2018, 06:29:09 PM <...> What you say is true. My proposed criteria is a good approach, but not necessarily the only one. Perhaps there could be quotas per (Top) Ranks in addition during the selection process. That is, select x Legendaries, y Heros, z Sr. Members for example, and work on the list in the same way but filtering previously by Rank to scout for candidates. Of course that play well if there are many candidates, but section/subsections with few candidates would not be able to play by the quota per rank criteria. I thought of also creating a variable with the size of the network that is unique (i.e. Merited by the given person, but not by anyone else), but thought it would be complicated to use and always as a side criteria in any case. Title: Re: User's sMerit Network size – Data to select possible Merit Sources Post by: Jet Cash on May 22, 2018, 08:19:22 AM Thanks for the kind words guys. I think I have been lucky in that I received a chunk of merits in the initial airdrop, and my policy of awarding them in ones and twos seems to have resulted in my picking up a few hundred sMerit in addition. In the main, I have ignored the rank of a poster when awarding merit, but I have tried to find threads where newer members were likely to post. I think I have awarded 3 first posts with a merit, and I did this where I got the impression that the poster could be encouraged to become a valuable contributing member.
Also, I gave Hilariousetc a merit for a single word post of "yes". I suspect that many members will consider this as a wasted merit, so I will post an explanation. I'm sure that a merit is unlikely to be life changing for Hilarious, but I believe that merit should be used to encourage good posting behaviour, and highlighting good examples is as important as rewarding the poster. I felt that the single word answered the OP's question completely, and further justification was not required in this case. The award was really because he avoided bulking out the post to try to achieve some minimum character count requirement. I hope that I can help to build Bitcoin Talk into an international community to discuss the new government backed 'cryptos', as well as Bitcoin as it makes its transition into a store of wealth. I started the translation initiative for personal reasons. I would like to hear opinions of the banking changes in countries such as Venezuela, Argentina, and Italy, and many of the information posts are confined to the local language boards. Bitcoin Talk is an interntionl community, and as such it has a few language problems. The bulk of the boards are reserved for English language posters, and I don't believe that this should change. This is for the reasons that hve been posted in many threads. I considered suggesting a multi-lingual board, but I rejected this idea. It would be a nightmare to moderate, and it would be very difficult to read any threads containing posts in a variety of languages. The translation of the better threads and posts from the local language boards seemed to be the best way for English language speakers to benefit from the knowledge and experience contained in the local language communities. I suggested that they are posted on the Ivory Tower, because I felt that most senior members would take a more responsible attitude towards thread selection. This is a generalistion, and I apologise to the competent and responsible members who are not yet able to post in the IT. Title: Re: User's sMerit Network size – Data to select possible Merit Sources Post by: seoincorporation on May 22, 2018, 03:02:04 PM This is a generalistion, and I apologise to the competent and responsible members who are not yet able to post in the IT. This is the best way, I think. If not, many newbies-merit-hunters will use Google TRanslator and spam such a great initiative. I believe you should postulate yourself to be a Merit Source, data has spoken and many of us agree with you are one of the best candidates. Of course, only if you want, jaja. But, given your recent initiatives with the local Spanish board, I am convinced: you should be a merit source. Title: Re: User's sMerit Network size – Data to select possible Merit Sources Post by: Jet Cash on May 22, 2018, 03:53:24 PM I believe you should postulate yourself to be a Merit Source, data has spoken and many of us agree with you are one of the best candidates. Thanks for the kind words, but I think that merit sources could be the victims of begging PMs. I'm also having trouble finding posts to merit. I think I need a cork screen for my computer, so that I can throw darts at it. Title: Re: User's sMerit Network size – Data to select possible Merit Sources Post by: Harlot on May 22, 2018, 04:45:47 PM Good idea, being generous to giving away their sMerits is one. However you are right identifying who is generous isn't enough to determine if that member is a potential merit source, we still need to check their merit history one by one and merited post by merited post as this is where you will really see if the sMerits they are sending are for high quality posts. In my own opinion this is really a good alternative of becoming a Merit Source for those who don't want to apply personally.
Title: Re: User's sMerit Network size – Data to select possible Merit Sources Post by: vlad230 on May 22, 2018, 06:55:19 PM These are very nice stats DdmrDdmr! Merited! ;)
Another usage of these stats could also be to track down merit abusers. Usually they have a very high sent merit average. Need to mention that some of the ones that have an average of >100 merits per transaction seem a bit suspicious to me if there isn't an error in there. I also found my name in the list too :) I'm the last one in your list though. ::) user_from name rank nSentTo sMeritSent nTx avgSent stdDevSent cVarSent NetworkGroup 1268116 vlad230 Full Member 31 32 32 1.00 0.00 0.00 (30..40] Not sure why my average sent is 0 here. I've only sent 1 merit per transaction so this should say 1, not 0.00 EDIT: I think the formatting is the culprit, it says 1 to my average. Title: Re: User's sMerit Network size – Data to select possible Merit Sources Post by: DdmrDdmr on May 22, 2018, 07:42:58 PM Good idea, being generous to giving away their sMerits is one. However you are right identifying who is generous isn't enough to determine if that member is a potential merit source, we still need to check their merit history one by one and merited post by merited post as this is where you will really see if the sMerits they are sending are for high quality posts. In my own opinion this is really a good alternative of becoming a Merit Source for those who don't want to apply personally. The only grey area remaining would be if theymos wanted to create Merit Sources on local boards that currently hold hardly any candidates, since some of them barely have any merited users to start with. That happens on around half of the local boards, where the sMerit network size threshold would be too low to be a reliable initial candidate selector. If there were to exist a will to create Merit Sources on such boards, options are low. Apart from volunteers postulated as candidates through the established process, the other option perhaps would be to initiate the rollout of sMerits through more trusted local people such as moderators (although not all boards have them nor do they have too much free time I guess). <…> Yes, that’s it. Your average is a round 1, with 0 standard deviation since 100% of your sMerit awarding transactions were all of the same value. I always struggle with the formatting of coded tables. I know that a table can be created, but it’s just so cumbersome.. It’s easier to see on one´s data on the general google doc file (you’re on row 219 of the global tab). You are right that this data could also be uses for merit abuser candidate tracking. The idea would be to filter the general file by high averages and (normally) small network sizes. Those are candidates. In my opinion though there rules to treating cases with merit abuse are not crystalline clear, as the assumption is that it will become rather meaningless in the overall picture over time. We’ll see. Regardless, tracking the cases down and resolving in negative trust is fine, but not always easy to prove and leads to uncomfortable situations both for the pursuer and for the pursued if the latter is a false positive. Title: Re: User's sMerit Network size – Data to select possible Merit Sources Post by: LoyceV on May 23, 2018, 06:08:34 AM I always struggle with the formatting of coded tables. I know that a table can be created, but it’s just so cumbersome.. If you'd make tables, you'd quickly run into post size limits (https://bitcointalk.org/index.php?topic=1782341.0). I think the easiest solution would be to fill out the right amount of spaces inside the code-tags.It’s easier to see on one´s data on the general google doc file (you’re on row 219 of the global tab). Title: Re: User's sMerit Network size – Data to select possible Merit Sources Post by: DdmrDdmr on May 23, 2018, 06:18:01 AM I think the easiest solution would be to fill out the right amount of spaces inside the code-tags. Cheers. Probably creating a the coded tag strings properly from my data environment should be easy. I’ll take a good look to get the hang of it. Seen the size limits too. edit 11:34AM I tried to place a larger list using two techniques: 1. Table -> Got the padding right, but length of post gets too long with a large list. 2. Code -> What I had before, but this time padded with spaces to align it all. The second one is probably better suited for large lists, but I encountered an issue just now: I placed 231 rows in the Code section, previewed the post, and everything look fine. Then I saved the post and it all went buggy. If cut half of the list off and also the bottom sections with text and graphs from the OP... panic !!!. Eventually I got the post from the draft history and altered it to show just the first 100 rows in two code blocks. Perhaps I encountered a max post length issue here, but it is weird since the preview worked fine. Anyway, I think I have not lost text at the bottom of the OP, so I'll leave it be now. It looks better, but for large lists it easier (and safer) to go to a google drive doc or such. Title: Re: User's sMerit Network size – Data to select possible Merit Sources Post by: DdmrDdmr on June 19, 2018, 04:40:34 PM Edit 19/06/2018:
Added this information to the Merit Dashboard (Link: BitcoinTalk Merit Dashboard (https://public.tableau.com/profile/ddmrddmr#!/vizhome/BitcointalkMeritDashboard/GlobalSummary)), under tab "Potential Merit Sources" so as to allow for a dynamic periodical update of the data, as well as to apply filters to it with recalculation of the ratios (by section/subsection, date period or specific user). Data on dashboard is up to friday 15/06/2018. https://i.imgur.com/ogKgSG2.jpg |