Semi-relevant, election modeling is pretty interesting and has a lot of similarities to derivative pricing.
Election Predictions as Martingales: An Arbitrage Approach
Why don't you write this in your own words? Scared of something?
Terrified.
Semi-relevant, election modeling is pretty interesting and has a lot of similarities to derivative pricing.
What would be the inputs to the model? Organized crime would need to be an input. Measuring the need for political change.
It's hard to say what a model needs or doesn't need to be accurate without doing a sensitivity analysis on parameters themselves, which I assume is possible with elections in aggregate given how many have and will take place.
Interestingly there is also variance in the sensitivity of the parameters so what's relevant one year may be less relevant the next. In the end the question comes back to how do you deal with uncertainty and how much uncertainty can be expected?
The point of the paper and video is that inherent to the system attempting to be measured, that is a system with binary outcome with unbounded uncertainty you cannot rely on a forecast as an actual estimate of the probability of the outcomes. This is because the forecasts themselves are random processes that will give a different result each time you run them. So to get an actual estimate you'd need to run them over and over again, which will yield something like a 50/50 outcome.
Here's an example I made for some crypto bros a while back.
These are both models of a coin flip, coin flips like forecasts are random processes.
The first graph is a coin clip that has low variance, so over all the flips the average outcome is always near the actual average outcome which is 1/2.
The second graph is the same thing but with more variance in the outcomes, hence the average outcome fluctuates around the real average by a huge margin.
Both will converge towards 1/2 eventually, the second one just takes forever to do so.
Now think back to 2016 when everyone was giving extreme odds in Hilarys favor, how were they so wrong? They don't realize their forecasts are random processes and predicting an election is useless given the amount of uncertainty, so they consistently get silly expectations.
In the future, if you see extreme outcomes being forecasted just realize the guys who make those models are retarded (Nick Silver)