Here is the reason this is a bad idea in as short a form as possible.
Fundamentally the question you are asking is a regression question. Given present measurements about the world [x], what is the predicted future state of the world [y].
The actual real world problem - the amount of information in [x], and the relative determinism of our reality in respect to [x], limits how accurate a prediction can be even with a perfect algorithm trained on an infinite number of prior examples of [x].
Essentially the problem you are trying to solve is a restatement of most machine learning problems. In most machine learning problems, the problem is in this form, and if you want to be methodical, you try out a large number of possible algorithms and tuning parameters to solve the problem.
At the end of the day, you do the following:
1. Feed the problem (you need many examples of past data) into an engine that will try to find the best fit model
2. Evaluate the outcome
If a good model exists, use it’s predictions to inform your decisions.
Either present methods can find a good model, or they cannot. Not every problem is solvable, and usually this is because it isn’t possible. We could have a prediction market on the next number a hardware RNG will spit out, but obviously no one can do better than chance.
At the end of the day the problem I have with your prompt is you posit that someone exists that can do better than just paying for the best tools (or professional actuarial analysis) you can afford. (even if the tool is free you pay with your time)
The exception to this is when there is hidden information. For example, if there were a prediction market bet on “there will be an assassination attempt on [VIP name] in the next week”. The problem here is the only individuals able to do better than chance are conspiring to commit the crime, and buying shares on the market would risk both alerting the security on the target, and revealing their real world identities and culpability.
Or “Apple will develop a car”. Same idea—the people who actually know are almost all direct employees of apple and buying shares on the market is just a form of being paid to leak information.
Here is the reason this is a bad idea in as short a form as possible.
Fundamentally the question you are asking is a regression question. Given present measurements about the world [x], what is the predicted future state of the world [y].
The actual real world problem - the amount of information in [x], and the relative determinism of our reality in respect to [x], limits how accurate a prediction can be even with a perfect algorithm trained on an infinite number of prior examples of [x].
Essentially the problem you are trying to solve is a restatement of most machine learning problems. In most machine learning problems, the problem is in this form, and if you want to be methodical, you try out a large number of possible algorithms and tuning parameters to solve the problem.
At the end of the day, you do the following:
1. Feed the problem (you need many examples of past data) into an engine that will try to find the best fit model
2. Evaluate the outcome
If a good model exists, use it’s predictions to inform your decisions.
Either present methods can find a good model, or they cannot. Not every problem is solvable, and usually this is because it isn’t possible. We could have a prediction market on the next number a hardware RNG will spit out, but obviously no one can do better than chance.
At the end of the day the problem I have with your prompt is you posit that someone exists that can do better than just paying for the best tools (or professional actuarial analysis) you can afford. (even if the tool is free you pay with your time)
The exception to this is when there is hidden information. For example, if there were a prediction market bet on “there will be an assassination attempt on [VIP name] in the next week”. The problem here is the only individuals able to do better than chance are conspiring to commit the crime, and buying shares on the market would risk both alerting the security on the target, and revealing their real world identities and culpability.
Or “Apple will develop a car”. Same idea—the people who actually know are almost all direct employees of apple and buying shares on the market is just a form of being paid to leak information.