Probability rules for how much longer something will last
If it’s a normal distribution, then use the average rule — something that’s close to the average, shifted a little bit by how much they’ve already been to so far.
If it’s a power-law distribution, then use the multiplicative rule: Take whatever sample data point you have, and multiple it by a constant value (for movies, it’s something like 1.4). Therefore, if you know that a movie has made 90 million, then it’s expected to make on the order of 126 million in total.
Finally, if it’s an Erlang distribution, the distribution modeling the length of the interval between two independent events, then use the additive rule: no matter how much time has passed, add a constant amount to predict how much longer it is going to last. This is a consequence of the memoryless properly.
For more practically applicable algorithms, see here 202006031201. As I fill out my book review for Algorithms to Live By 202005302204, I’ll add the things I learn from there into a broader list, that contains this note inside of it.
uid: 202006022154 tags: #knowledge