RE: A new approach to Content Reward Allocation
Here's the kernel of a simpler alternative, which captures some of the prediction market component without needing blinding, and also without needing a complex UI.
Post score is a number from 0 to 1 which changes as votes are cast. Perhaps this could be displayed as a "star" or numeric 1-10 rating (tbd, and this aspect could be split tested). Each vote is viewed as a prediction of whether the future post score will be higher (upvote) or lower (downvote) than the current score at the time the vote is placed. Votes are evaluated for payouts on the basis of how much the post score changes between the time the vote is cast and the end of the scoring period (24 hours?). Total payout is some combination (tbd) of resulting score and vote weight cast.
This provides an incentive to downvote, if the post is getting too many early upvotes relative to its more thoughtfully perceived quality, or vice versa.
EDIT: One detail I omitted is that every post starts with a prior average of 0.5 with some predetermined small but nonzero weight. This provides a basis for evaluating even the first vote, and it also provides a mechanism to continue to give some very small weight to later votes that endorse a strong upvote (because the average will never quite reach 1, even if all votes are up).
Ohhhh...ahhhh...I like that! (You are a smart cookie)
I also like this idea.
Wish my votes actually pushed so I could get this towards the top. This is good.
Buy Steem Power :)
Sounds cool
Need to figure out a way to reward the ones who down-voted a "bad" post especially if it ever has a high score, make curation rewards less related to the post's final score.
Yes I agree. But there is really no reason why the post score and vote score have to be related at all right? They're just coming out of two different pools effectively.
After thought deeper, I found there is an issue with this approach: for example, if current score of a post is 0.55, a whale thinks it should get higher score for example 0.6, so she up-voted, but perhaps the score become 0.9 after the vote and hard to correct unless she revoke the vote. From this POV, dan's proposal is better: vote on the result but not direction.
This could be done under something close to the current mechanism by using reduced vote power (indeed I already do this sometimes when voting for content that is good but not great). I think that's pretty much what it means for a whale to upvote something but still believe it is really only good enough to rate slightly better than average. A full power vote would also tend to give such content a large reward, which makes little sense.
There are some other interesting aspects to this, and I may write a separate post if I have time later.