1. IntroductionBack in December 2018, I analysed the Merit/post ratio for all the main first-level boards (see
Analysis – Merit per post per Section/Subsection). I wasn’t really planning on updating the information, but I did kind of say that I’d try to at some point when answering some PMs, and then additionally I saw a comment by o_e_l_e_o in the same line.
Averages change over time, and bearing that in mind, what I’ve now analysed covers the period
26/12/2018 .. 27/02/2020. I may at some point extend the data to previous dates, but if I do, it will only be for the Merit Dashboard.
The data is derived from:
- The aggregate of Merit TXs files.
- My scraping and classification of merited posts.
- The number of posts, as per snapshots on TheWayBackMachine.
Considerations:- Since the number of posts is derived from TheWayBackMachine, I do not have a choice in the dates that are snapshotted there, thus the apparently weird date ranges. I try to make them as close as possible to a complete month, but the snapshots are simply not as one would wish.
- The post count number may differ slightly from that on the Post Count tab for local boards. This is due to some chosen periods of time now differing a day or so, and the fact that I may not have used the same snapshot on both (some dates have multiple snapshots).
- The number of post is the Net number of posts created between two snapshot dates. Net means: Created - Deleted – Moved +/- Reindex Adjustments. Most data points are consistent and logical with other month’s data, but some may be rather different. Averages may shift wildly one month due to post deletion, movement or reindexing.
-
Averages are an indicator, but can also be deceiving (specially if post count varies). Because of the previous point pointed out, it is best to look not just at the summarized average merit/post, but also at the individual values over the past months.
For example, the below table states that the average merit/post on the Spanish Local board over the last 14 months is -0,07 merits/post. That seems wrong, and as an aggregate it is. What went on behind the scenes is that something like 51K posts were deleted/moved/reindex in between the 24/02/2019 and the 26/03/2019. This greatly impacts the calculus of the total merit/post ratio. It is best to look at the individual monthly averages there, and exclude the abnormal month. This can be done on the Merit Dashboard (which would place the average on 0,27 merits/post). Likewise, the 0,82 merit/post ratio last month was also due to post deletion/movement/reindex.
- Averages do not explain the story, and sometimes we really should dig down into the context of the data (i.e. massive post deletion), and even into the underlying childboards. For example, most of the local boards have a better merit/post ratio now than when I performed this exercise at the end of 2018. What’s changed? Well:
-- Merit Source reallocation
-- Large Decrease in the numer of posts, with émphasis on Altcoin/Ann sections
Those two factors alone shift averages quite a bit on local boards, so comparing now to then needs this context in mind.
In addition, WO is behind a great part of the merits in the Economics section. It would be very interesting to do this sort of analysis distinguishing between boards and childboards within, but that is an overkill task to perform.
Note: There is a fair share of manual work behind the scenes to retrieve the post count numbers. Though I am careful, I can make mistakes. If there is any specific weird figure that you’re interested in, I can revise it and try to bring a shed of light on its nature.
2. Merit per post
The above table summarized the average merit/post for the last 4 months of data I have. The lasts columns display the aggregate merit/post data
for the period 26/12/2018 .. 27/02/2020.
When browsing the data, I’d check all 5 merit/post averages on each row. Some are pretty consistent such as Beginners & Help, but others need context such as the previously mentioned Spanish board and Meta (Meta is really more on the 0,7 Merits/post than the 0,93 displayed – the Art Contest drove the average to 4,15 merits/posts for a month).
Just to reiterate, sharp movements in averages have a contextual reason behind that the numbers do not tell.
3. Merit per post on the Merit DashboardEdit: There seems to be some issue visualizing on Tableau the content of the filters on the two below added tabs. I've never seen that happen before, and I'll need some time to see what the issue is (Ipad ok, pc empty filter combos). Values as there, but on a PC show in a tiny font size.I’ve added the above table, with all the full 14 months of data to the Merit Dashboard (last 2 Tabs). There, you can play around with the filters to:
- Delimit the comparison to a set of specific boards.
- Delimit the months (i.e. exclude abnormal values).
The output is something like this:
https://public.tableau.com/views/BitcointalkMeritDashboard/MeritperPost?:display_count=y&:origin=viz_share_link
https://public.tableau.com/views/BitcointalkMeritDashboard/MeritperPostIIChart?:display_count=y&:origin=viz_share_linkAnd yet again …
values need context. The Spanish 0,82 reading is due to a number of deleted/moved/reindex posts (I personally would take that month’s reading as void when interpreting the data).
Well done on the study.
