This week sees some significant changes to the data structure of the report, which makes comparison to prior weeks more difficult, owing to the fact that there is an enriched data set now in use.

I have also got some new features in the data, in the interest of continually improving the veracity of the data.

Whats New since last week?

---

I have now, for the first time, got Leo Threads / InLeo data working in the data set. Leo Threads has a lot of activity, but as far as I am concerned, it isn't really a post, but stuff published via InLeo is.

Therefore, including InLeo data is done by the following methodology:

IF POST by application LEOTHREADS and GREATER THAN 250 Characters INCLUDE, ELSE EXCLUDE

(In simple language, that hopefully anyone can understand)

Furthermore, there are a number of accounts that make "posts" to the blockchain, that aren't really posts, so I am going to show the list of accounts that is excluded for the purpose of displaying the remaining data that is analysed in the rest of the post.

The current exclusion list of accounts are:

• ai-summaries
• waivio.guest03
• guest06
• guest07
• guest08
• guest09
• hbd.funder
• peak.snaps
• ecency.waves
• indiaunited
• lolzbot
• guest10
• waivio.guest04
• pizzabot
• buildawhale
• leothreads

If there are other accounts that you think should be excluded from being counted as "posts", please let me know, and I will take a look to include them in the exclusion list.

Anyway, onto the usual reporting:

For the week in question, there was $40,318.21 in rewards distributed. The average payout was $2.01 The max decreased from $118.41 to $83.82. The median payout increased by a single cent to 63 cents.

The highest word count was 10,637 (again!) (is there a duplicate post, or is it a coincidence?!), the average was up to 469 words and the median increased from 273 to 296.

Here we can see the distribution of posts by word count, and the percentages that of all content by word count. Also, below, the sum of words written on chain in each category.

Pay per word is a little bit lower, from 0.0048 to 0.0043

This week, a long post got the highest payout! Congratulations to curators! This is an excellent sign of the fact that things are getting rewarded for longer form text. Still, the bulk of rewards go to content under 750 words, as per the pie chart above. On the whole, the average payout for posts under 1500 words (but more than a thousand words) seemed to get the highest pay out. It seems like that might be the sweet spot.

I plan to expand this to look at images per post, too, to get a better understanding of differences between posts that contain a lot of pictures, and posts that contain a lot of words, and vice versa. There is also some problems in assessing content like this as much content is published with multiple languages in the post, inflating the word count.

Typical Posts

What about the typical post on HIVE? What does that look like? Well, thankfully, here's some stats - this week with some new data points, average replies!

This week is sort of a new base line, given the inclusion of InLeo data for the first time, but it hasn't seemed to impact the numbers on the whole all that much.

Payouts

How does total, aggregated payout look, compared to the word count? A little bit like the distribution of the posts, it seems, with a steep falloff for longer posts.

Shorter posts continued to get "more" rewards proportionately. A good uptick in the 501-750 words category, and a reduction in rewards for longer posts under 2000 words seemed to continue, as a percentage of the total.

This Week

Posts between 251 and 750 words got the most rewards. There's a pretty big fall off after that point.

Reward Factor and Other Stats

How about when we consider a few more variables, and look at the reward factor? That is to say, the highest payout in a word count group divided by the average?

The reward factor shrunk by 14. That means the highest rewarded posts are taking 41.75x more than the average payout across all the word count categories.

Moving On

40% of all posts published on Hive were above average word count and above average pay out. 12% were below average word count, and above average payout. There is room for awards to be allocated by curators to those appearing in the above average word count and below average payout category. There's ~$6.5k of awards allocated to shorter posts.

There reward factor for "below average word count and above average pay out" grew again this week sharply - an increase from 19.59 to 24.35.

That's max payout divided by average, presented as a number.

What about when we look at authors on an individual basis, for the period? (Or at least the top 20 for each category?

I have added engagement, by measure of total number of replies to an author's top level posts here. I am curious as to the range of comments received by authors in these categories. There's a wide variation in the engagement each author gets for their rewards, regardless of their word count.

The same authors tend to appear in the lists, week on week. They're consistent and persistent.

What about those potentially "emerging" authors, with Below Average payouts?

Different authors are continue to appear in these lists.

Finally, I have put together a view of the key metrics I'm looking into for this data set into a slice by community, sorted by total payouts.

We can now see LeoFinance data (No one told me it was missing in any of the previous reports, as now InLeo posts are included!) And Splinterlands sees a return to the top 5, on the basis that the new free to play mode was released during the period, seeing an uptick in posts on the category. Spend HBD is down the list, and Photography Lovers is significantly down the list.

I am still planning on doing a post focusing on the photography communities of HIVE with the new data set, but I am not sure when that will be.

Who the fuck swears the most in their posts on HIVE?

The only words, as an Aussie, that I consider swearing are "shit", "fuck", "cunt" (or variations thereof, so this captures "fucker", "fuckhead", "motherfucker", "shithead", etc

@galenkp, regains his swearing crown. I remain in the list, probably by virtue of this very weekly post, and the paragraph above this one. I find an interesting distinction between the list of users that swear also being the list of people who I often do not miss actually reading a post of - I see @harbiter, @riverflows, @honeydue, and @blanchy all in this list, and these are users that I will always have time for. I always read their posts because they are engaging, funny, deep, or a combination of the three.

I also generally read the content of @tarazkp and @galenkp, as they have great vocabularies, and interesting posts.

Conclusion

Thanks for taking a look at the data. I'm still keen to get more ideas from you all on how this can be improved for future iterations and hope you enjoyed the new features, or whatever other metrics I might be able to extract.

Check out my other recent posts in the HIVE Statistics community to get a guage on the previous discussions these posts have raised.

Ultimately, the questions I am asking by presenting this data is to curators:

Are you rewarding content appropriately, and are you letting autovoters distribute rewards the same way each week, instead of hunting for content?

My suggestion is that you curate manually, enjoy the content and engage, as that is where the strongest relationships are forged on HIVE!