Prediction Markets Are Flawed In Theory, Not Just In Practice.
In the limit, market capital is dominated by the forecaster with the highest long-run growth rate of capital. Because this rate is a function of both accuracy and risk management (calibration), there is no guarantee about which point on the accuracy-calibration curve the market will stabilize at.
The long run growth rate of capital depends on calibration and accuracy over a series of bets. For any finite subset of bets, there may be forecasters with arbitrarily superior accuracy or calibration compared to the market's dominant forecaster. Nevertheless, their influence on the market price will be washed out by the massive bankroll of the dominant forecaster. The frequency with which the market presents questions on which a trader has an edge is as important to determining the dominant forecaster as calibration and accuracy.
Per the Grossman-Stiglitz paradox, markets require a sufficient level of inaccuracy to motivate traders to participate. See also the No Trade Theorem. This is nonfatal for the stock market, which is primarily intended as a financing mechanism for productive economic activity. It is a serious flaw for prediction markets intended as a forecasting mechanism.