Proposal: Game where I simulate polls and you try to predict "elections"

[Nichlemn]

I notice in threads discussing opinion polls, there's a lot of talk about "junk polls", "junk pollsters", and predictions where people put a lot  of weighting on their favourite polls. I was thinking that it could be interesting if the bad pollsters and true support levels were knowable, so we could test how good you are at evaluating polls and making predictions.

The way I'm thinking it would work: there's a two party system and polls measure a simple Party A or Party B vote. (I could try make up a back story to make it more interesting). The support level for each party changes every month (in the story, in reality each post could have many months of polls), by a random number generator. Then, I simulate a whole bunch of pollsters "polling" each time, with varying degrees and types of bias. Maybe some systematically favour one party, maybe some systematically overestimate swings, maybe some are just made up manually by me (representing a fraudulent pollster). Every so often, there's an "election" you have to predict. After a modest number of elections, the next exercise is to guess how each of the pollsters is biased.

[Cranberry]

I would be interested. Sounds like fun

[Gustaf]

That sounds like an awesome idea.