I made some calculations and it turns out that the <1% answers significantly reduce the effective liquidity of the remaining answers.

[EDIT: this is wrong because I was mistaken about how the AMM works for multi-choice questions]

So actually, I did some tests and it seems the manifold market maker doesn't follow the constant product rule for multiple choice questions, so the calculation above might be wrong. Looks like I will have to dig into manifold source code.

@FH7979e there are some specific edge cases like very small pools are omitted. (Or were at one point)

@Eliza the exact mechanism of the buying and selling has a good code comment, about how they first buy excess shares of the other type then use the leftover to buy more shares and iterate that loop until there's none left.

@Eliza As I understand it now, the mechanism for multi choice questions is that there is a separate binary CPMM for each answer, and the system automatically arbitrages between them. This means that each answer has its own separate liquidity pool. You can see this in the API - e.g. https://api.manifold.markets/v0/slug/who-will-be-elected-president-in-20 has a separate pool for each answer, and each answer has a totalLiquidity. What's not yet clear to me is how is liquidity distributed among the answers when it is added to the market.

When someone adds 1000 liquidity to a market like this and it is divided among all the 0% answers, does it have less than 1000 liquidity worth of effect on the market?