However for most users the chances of a large account coming along and upvoting their post will remain small, even if the shift in incentives does change behaviour
There is more to it than that. An enormous portion of the reward pool is being milked out by self-voting and vote selling. Even a partial counter of that will increase the value paid out in STEEM per vshare. That also offsets the lower vshares per rshares at the low level of the curve resulting in higher rate of STEEM payout for those fewer vshares at the low end, even if no large stakeholder comes along and votes.
There are too many moving parts to do a simple calculation of payout for a particular vote assuming all else is equal and expect that to be remotely valid. It isn't.
It's true that significant use of downvotes on bad actors would cause an upscaling of rewards on all other content. It's something I should have added into my estimates. However the impact of the SPS / 50/50 and CLRC all act in the opposite direction and those impacts will dominate heavily for small earning posts.
Here I would respectfully disagree. Most of the proposed changes have clear and measurable impacts and getting an initial assessment prior to behavioural changes is useful and a pretty solid starting point. People can then judge how they think behaviour will change those figures.
We can respectfully agree to disagree. A large portion of (in fact nearly the entirety of) the intent of this is to significantly change behavior. So I don't think that an assessment prior to behavioral changes is a solid starting point in the sense of being a good estimate of the outcomes. It literally can be a 'starting point' in the sense that one can start there and then make large adjustments to get from there to a good estimate, which is what I am pointing out.
It is also quite possible IMO that EIP is a complete failure in the sense that there aren't major behavioral changes, but that's somewhat of a different claim to make.