As you may remember, in these two posts in I criticized the way how 'curation' currently works on HIVE (and before on STEEM):
I think one essential problem is that most of the time curators don't read what they upvote (so there isn't a really high correlation between the number of upvotes and the quality of the content), and ...
... one of the main reasons for that is the small 'curation window'.
Summary of what bothers me
As I wrote in these articles I think we should remove the five minutes curation window (directly after a post was written), because it clearly disadvantages manual curators who try to read and evaluate posts before they decide to upvote them or not. Before they could do that, most of the time lots of automatic upvotes have already come in.
It also makes 'curators' lazy (especially since there is 50 % curation reward): they simply upvote the same people again and again (there even aren't 'diminishing returns' implemented), while not enough users are seeking for new authors (respectively honor posts from old authors who aren't on their 'lists').
I see no reason to be rewarded less for a late manual upvote than for an early automated upvote!
And considering the high percentage of automated upvotes, I think it's simply a myth that manual curators find 'quality content' first, and then more upvotes follow and reward them (yes, that may happen ... as an execption).
New idea to improve curation
Recently I was thinking about how to make it less attractive to just upvote early without reading anything, and my conclusion is that curation reward should solely depend on the weight of an upvote, neither influenced by the date of the upvote nor by the (expected) number and strength of upvotes of other users (which currently can be a reason to decide either in favour or against upvoting a post). Curation rewards would be only determined by the available HP of the curators and the percentages of their upvotes.
"But how does that fit together with the convergent linear rewards curve, @jaki01?"
Well, I like the current convergent linear rewards curve (actually I suggested something similar long time before it finally had been introduced), so I think as curve for the author rewards we could stick with convergent linear, but for curation rewards the simple formula that curation rewards would only depend on the vote weight of the curator could apply!
Then there was no reason anymore to upvote certain users early just because of expecting other upvotes to follow (the same applies for the or the opposite case, not to upvote a post when it has already collected too many upvotes), while great authors still would benefit from getting many (strong) upvotes because of the convergent linear rewards curve which still applied for author rewards.
Disadvantage of my idea
It's true that self-voting would be somwhat more beneficial again as self-voters would receive more curation rewards than before. At least author rewards would still be limited by the convergent linear curve (and don't forget the 'free' flags since the introduction of EIP ...).
In case this bothers you too much, I offer you another suggestion which, however, is quite an attack on EIP:
when going back to 75 % author rewards (and 25 % curation rewards) the increased curation rewards for self-voting described above would decrease again ... But well, I better stop here, not to trigger another flag war against or in favour of EIP. :-)
Convergent linear curve for author rewards (no change needed).
Curation rewards depend only on the own vote weight (HP, percentage of the vote).
Possible addition (if self-voting is considered as serious problem because of 2.): going back to 75 % author and 25 % curation rewards. :-)