Title: Analysis – Personal merit per day, merit per post and ranking-up projections Post by: DdmrDdmr on October 03, 2018, 02:33:37 PM 1. Introduction.
I was curious as to what the merit per day and merit per post for a forum member looked like these days, and what a projection of their ranking-up dates could roughly look like. While the merit per day indicator is relatively easy to derive, the merit per post and the projection of the dates when one is to rank-up are not. I’ve tried a few approaches and posted here the one that seems better off out of the diverse alternatives I tried. 2. Earned Merit per day. The earned merit per day for a given person is calculated as the coefficient EarnedMerit/nDaysReference, where: EarnedMerit = the sum of all non-airdropped merits. nDaysReference = The number of days between 24/01/2018 until the 28/09/2018 (last data extraction-> early morning-> virtually as if it were the 27/09/2018) that the person has had his account for. Accounts created on or before 24/01/2018 will all have 248 days (number of days the merit system has been running until last data extraction). Newer account will have a smaller value logically (i.e. an account created a month ago will have 30 days in this parameter). Upon further though, I saw that even this simple ratio is not always fair. For example, let’s say we have an account that has been created 1 day ago, and received 15 (suspicious) sMerits. That account would have a EarnedMeritPerDay ration of 15, which would be the top ratio. For that reason, even though I’ve calculated internally the indicator for all accounts, I would omit from any ranking those accounts that have not posted at least 30 posts recently (within the last 4 months *), which seems to cut out most of the extreme cases I had cropping up in the top places. Top 25 EarnedMeritPerDay: Code: user_id name rank EarnedMeritPerDay earnedMerit nDaysAcc. nDaysReference url • Satoshi is off the list due to not meeting the 30 recent posts criteria. Another great ratio that is chopped off is @theyoungmillionaire (his ratio is 2,23) , for the same reason. • What surprised me is that, out of the 21.971 all-time merited users, 11.424 has a post count of less than 30 post in the last four months (*). It is not just the Newbies and Jr. Members, since there are plenty of members, full members and other ranks in the same situation. That indicates that just over half of merited people post rather seldom or in bursts. • The full Merit user base distribution for EarnedMeritPerDay is as follows: Code: EarnedMeritPerDay nUsers % (*) Nº of Recent post within four months is calculated as follows: - If the account is < 4 months old then total posts are considered. - If the account is >= 4 months old and account is in backup from four months ago then total nPostNow – nPostBackup - If the account is >= 4 months old and account is not in backup from four months ago (merited in between for the first time) then (total nPostNow /nDaysAccount)*nDaysIn4Months. Basically, using all-time average per day for the account, and translating that into 4 months. This is not perfect fot the third case, but it is a decent solution for the most. Nevertheless, there are circumstances that shift the post count form what it really should be in reality, such as deleted posts. 3. Earned Merit per post (last 4 months) The earned merit per post for a given person is more complicated. The ideal would be to have a full user snapshot of the DB from the 24/01/2018 to contrast current posts against. Ideally, all posts prior to that date should not be considered (although arguably backtracked posts were merited in the early days of the Merit System). I also thought that after eight months, people’s received merit patterns may vary rather, especially if we contrast the early months to the rest. In the end, I decided to use the 4 month criteria for this calculation as explained above for posts (*), performing a similar calculus for the number of merits received in the same timeframe (consider all earned merit if account < =4 months old, merit difference from 4 month backup for those in the backup, average of up to 248 days for the rest and multiplied by 4 months). With the above, the top 25 users with most earned merit per post are: Code: user_id name rank earnedMeritPerPost nMeritReferenceCalculation nPostReferenceCalculation url 4. Ranking-up dates I also gave it a go at calculating the rank-up dates for forum members that had PostsPerDay>0. Looking at the distribution shown above on merit per day distribution, the majority of the projections are bound to be really far ahead (I mean 91,55% of merited users are being awarded below 0.1 merits per day, so for most, the stopper is merit and not activity in general, as we already know). Basically, I’ve played around with the MeritPerDay ratio and the PostPerDay ratio. Then comparing with each rank’s merit and activity requirement (the minimum 775 for Legendary), I calculated the number of days to get to each rank on account of both concepts. Finally, the date is derived from the maximum number of days of account of both tolls (which ever takes longer to comply- the merit toll or the activity toll). This is really a rough approximate, and not a solid one-to-one forecast, since past month's data is used to project behaviour in a flat linear pattern (and we know that merit generally tends to decrease really). However, the exercise is interesting to perform. Projections - Additional people that will rank-up with the Merit System: Code: year Members Full Members Sr. Members Heroes Legendries Note: Some member’s projections are clustered rather largely, since they have the same amount of merit. For example, the 3.950 Member cluster tend to have currently 1 or 2 merits, many with activity already ripe for Member, but not merits. Also I grouped years to make it more readable from 2020 onwards. For example, the 4 Legendries for 2018 (reaching at least the minimum activity level + 1k merits) would be @Hhampuz, @BTCforJoe, @HCP, and @marlboroza (**). For 2019 we would have as forecasted Legendries: @bill gator, @bob123, @buwaytress, @coinlocket$ (end of 2019), @gawlea, @HairyMaclairy, @HeRetiK, @Iasenko, @joniboini, @killyou72, @krogothmanhattan, @LeGaulois, @Matthias9515, @Micgoossens, @mstfprcn, @o_e_l_e_o, @PHI1618, @Piggy, @sabotag3x, @stompix, @TheQuin, @Tonych, @TryNinja, @xtraelv. Note: (**) Aside, there are also currently 85 Heroes with enough activity (at least 775) and merit to rank-up at some point too. |