/biz/ - Business & Finance

>>49374817
You would just end up roughly averaging out the 5 machines
I recommend the same, but you track the machine’s odds and only switch if the machine gets a loss that puts its overall odds below a different machine

Slight improvement could be made with simulated annealing in case your first pull of the 55 is a loss

Approach 2:
Do a quick population sample test of the 3 machines, with something like 20 pulls each to “prime” their odds (the expense of this is negligible in the context of 100,000 pulls)
See also: Optimal Stopping problem
https://en.m.wikipedia.org/wiki/Optimal_stopping
There’s an example of this involving a Saudi prince that has to pick the prettiest wife from a line-up of 10,000 women, but I can’t find the link to what it’s called
>pic related to this approach