There's at least one missing factor in your idea and that's what is the impact of the voting power used by the curator on his chances to win, so this would need to be worked out, but I guess there's some obvious ways to do that (although they might make the whole idea less appealing).
Also, I'm not sure that vote stacking is really profitable today for curation rewards,but I think most people just aren't aware of that (or just don't care enough about their rather minimal curation rewards enough to modify their behavior).
Your idea is interesting, but it's a bit complicated and should probably be given more consideration after trying some simpler tweaks first. The gamification effect of a winner-take-all system might encourage people to care more about the reward, but only if it didn't always seem to end up in the hands of a whale.
So, I forgot to consider that. In order to achieve the effects of decentralization and avoid spamming the network with small .1% votes, you would probably have to use VP% as another weighting factor when calculating odds.
One could also consider adding the actual voting strength to the calculation, but I worry that this would not be able to achieve the decentralization of rewards given the massive disparity between the votes of minnows and the votes of whales.
The more I think about this, the more I fear it is flawed a concept given that there will always be those who gamify any system as soon as they find a loophole.
The loophole here is that somebody will just create a farm of upvote accounts. And of course, there’s voting trails. Long term a whale will not be interested in discovering the good stuff anymore if they see that their search for quality content written by less popular authors (sorry, there’s so much not popular enough quality content on Steem that the game isn’t all about minnows) is constantly rewarded by zero because of voting trails diluting their possible chance to win. Which seems a core idea in the reasoning behind this idea.
Why wouldn’t a whale just delegate away then? No matter whether for profit or to other curators.
Make it a game, fair play. But don’t overreason it because the more you try to reason things, the more obvious loopholes become. This won’t fix voting guilds/trails, only if the original upvoter gets a higher chance to win. But that’s hard to determine. Not every app can be obliged to open up an API, and any pattern recognition algo may be too resources hungry. Whales will still have their trail because they still open up the highest rewards, so the argument that they will want to find less popular quality content because of less competition is an oxymoron, almost.
Of course, there’s always the option to devaluate the value, and rewards, of automated votes by implementing a dual citizenship structure in which apps must identifying as such,. Failure to do so may result in a whole account (with whole wallet) being frozen, locked, and eventually deactivated.
To be honest, the evolution to such dual citizenship structure is the only way to ever possibly start tackling whichever form of automation is loathed. At the same time it would also introduce an option to completely freeze an account, something which as side effect may become handy a tool for bodies such as steemcleaners. But that also requires rules.