You may want a system where, the more skewed a bet is, the less money gets paid out by the majority side and more money gets paid to the minority. That way, bets like those don't really get you anything, unless you bet the unlikely case and win, then you get a well-deserved jackpot.
If you divide by ratio, and add a fixed number (e.g. 2) to the "for" and "against", there aren't really false positives, and true negatives (e.g. a reasonable bet where people just happen to all bet one side, or a biased one that doesn't get much attention) are unlikely - unless there are too many bets or people are gaming the system, that is.
This pari-mutuel system (I always knew the system but did not know the name) seems near canonical to me. And the incentives seem clear cut. It seems to me that if you feel the odds for or against any prediction are wrongly calculated, then if you submit money to push those odds towards the correct ratio, then you should, on average, expect a return.
... I maybe should actually run simulations to check that, that's just what my intuition is telling me as to how it would work.
I'm trying to understand what you are proposing and the motivations for it.
By ratio, I assume you mean the ratio of the pool for and the pool against? So if prop A has $100 and prop B has $10, then the ratio would be 110/10, or 11:1. So your procedure would be to add a fixed number to this? Either 112/12 or 13:3? or perhaps you meant Prop A $100/(110/10)+2 and Prop B $10(110/10)+2
So Prop A is ~11.1
Prop B is ~2.9
This may or may not be a problem, but I've found it requires more... infrastructure (notifications etc) for people to avoid being confused by it, and give people a chance to withdraw their bet if the odds are no longer favourable (alternatively put in stops, but that also complicates the system.)
https://news.manifold.markets/p/above-the-fold-market-mechan...
I wish I could post a link, but in their Discord they had some really interesting chats and I think they pretty much invented a novel AMM for this. They are open source, so you can take a look here: https://github.com/manifoldmarkets/manifold
I forget whereabouts in that repo the algorithm is though but have a hunt!
Happy to chat more about prediction markets/AMMs, feel free to reach out to austin@manifold.markets