Upvoted for interesting idea, but I don’t think it’s wise. A comparison to standard probability will demonstrate why:
You have a jar with 40 red balls, 20 white, 20 green, 10 blue, and 10 yellow. What strategy maximizes your chance of predicting the color of ball you get?
predict red, always.
randomize − 40% to predict red, 20% green, etc.
Strategy 1 has a 40% chance to win. Strategy 2 has a 26% chance to win. I think some of your “good” list should be on the bad side, and you’ve missed the most important selector: most likely to be correct (whatever “correct” means to you, using the same definition that you assigned probabilities in the first place).
You also have a very tough problem of enumeration and assignment of probability. That tail is pretty darned long.
Upvoted for interesting idea, but I don’t think it’s wise. A comparison to standard probability will demonstrate why:
You have a jar with 40 red balls, 20 white, 20 green, 10 blue, and 10 yellow. What strategy maximizes your chance of predicting the color of ball you get?
predict red, always.
randomize − 40% to predict red, 20% green, etc.
Strategy 1 has a 40% chance to win. Strategy 2 has a 26% chance to win. I think some of your “good” list should be on the bad side, and you’ve missed the most important selector: most likely to be correct (whatever “correct” means to you, using the same definition that you assigned probabilities in the first place).
You also have a very tough problem of enumeration and assignment of probability. That tail is pretty darned long.