Synthesis of debate between experts using LLMs to generate predictions

A group of people would be gathered, mostly experts in a field we're trying to predict. For example when predicting soccer match between Barcelona and Manchester City, various football experts, fans, bettors, even statisticians could be there. 

They would chat about the upcoming match or anything we want to predict to predict its result. The chat would be probably done with a moderator so it's usefulness would be ensured. We need a chat full of arguments and reacting to each other, without personal attacks or shallowness. 

Next, this conversation history would be fed into a LLM that would judge the arguments and make the prediction. 

Why this could work: 

1. wisdom of the crowd is a successful concept of predicting things
2. variability of opinions and arguments in the data
3. variability of data sources in the data
4. variability of subjective context in the data

And most importantly, an amazing new data would be created that is very hard to get elsewhere. 

Members who the LLM considers the most important for the prediction might get some money.

This could be scaled into a Numerai-like crypto project (need a token for participation, staking the token, being paid in the token)