A potentially touchy subject in light of recent discussions, but there's plenty of data and I think some positive outcomes to focus on with regards to curation rewards.
The discussion is mostly based around the chart above which uses the previous 10 weeks of numbers from the Curation Leagues top 100 earners, as seen most recently here.
I'm using the number calculated by dividing weekly curation earnings by total Steem Power held, and then multiplying that by 1000 to get the SP earned per 1000 SP staked.
e.g. If you have 1500 SP vested and earn 3 SP a week in curation rewards, your weekly earnings per 1000 SP vested is 2 SP.
The number has been calculated for each of the top 100 over 10 weeks, and a couple of math formulas have been applied to produce the chart.
In the chart there are just over 6 weeks of earnings tallied prior to the (first) hard fork on the 27th August. During this time, as you would expect, there is little movement in the average earnings - pretty flat at around 2 SP per 1000 SP staked. The low 20 average remained pretty close to 1 SP per week, but did tick up a little to 1.18 the week prior to the fork, and the top 20 average also remained quite steady, finishing up at 3.57 on the 25th August.
The weekly results following the 27th (in which we had around 5 days worth of collection under the new 'rules'), curation rewards ticked up across the board. This was generally expected due to the author/curator split moving from 75/25 to 50/50, but what else does the chart show?
Not long after I started the curation league, it was clear that those heading the list were using a combination of auto-votes and, either manually or programmatically, front-running bid-bots. The former I'm not against as long as it's not whole stake going to the same 10 accounts each day, and I repent my auto-voting sins by trailing 
The front running of bid-bots though I've never particularly liked. For almost two years, this is where the best curation rewards have been acquired, completely stuffing up any proof of brain and desire to curate and hunt for under-valued content which could gain traction and votes, earning the early supporters a nice reward.
However, this has now changed somewhat with the new 'rules', particularly the introduction of 2.5 free downvotes and I think the chart above represents this in a couple of ways. I mentioned above that scores ticked up immediately following the fork, the low 20 average moving to over 1.5, and the upper 20 average to almost 5 SP per 1000 owned - and then what happened? The low 20 average has continued to rise, but perhaps more notably, the top 20 average has fallen and is now lower than pre-fork totals, even with curators receiving double from content payouts.
The results on the 2nd September were the last week where unfavorable voting was economically much (4x Traf/Kevin?) more viable than favorable voting. In the weeks that have followed, those front running the bots, or placing a vote on content they knew (with 20+ months of data showing this) would pick up the same large votes, have been losing out on their rewards, whilst the general population have been earning more.
The top 100 average is up almost 42% from the week prior to the fork to yesterday, but even more impressive is the low 20 average - up 56% between the same dates. And if we look at the top 20 average and low 20 averages on the 25th August and 15th September, the difference is now 72% from 280%.
I think that's a pretty good result - there are plenty of manual curators in the lower 20 average range now earning more SP each week, while those voting indifferently are, at least at present, earning a little less. Promising signs.
Cheers
Asher