weka511
commented
6 years ago Sean, How does this sound? Agent: if you move to the high pool, your profit will be $xxx. Me: sounds good if you can be believed. Should I risk $tau on your advice? Flicks through mychoices to see how many times I went in that pool, then though mypayoffs and estimates return. Now it is easy (unless I have no history) to calculate the return in the past.
Now I compare to tau: either treat it as a bar on a rugby goalpost (no score unless the ball goes over the bar), or make a random choice, with probabilities (return,tau) (with obvious scaling).
What if there is no history, i.e. what if I have no experience of this pool? Matthew posted a link to the multi-armed bandit problem 2 weeks ago, which I repeat here. Maybe we follow a similar "epsolin greedy" approach to assign a credibility to the prediction.
Did you know there is a brokerage in Sydney called Fat Prophets?
E.g. if we are near the end of the game, and tau exceeds likely payoff