hot->warm->cold isn't primarily controlled by time.
hot->warm is mostly controlled by number of open hot buckets and the size of the hot bucket.
warm->cold is mostly controlled by size of your warm volume and number of warm buckets.
Data retention as in "how long to keep data in an index" depends mostly on three things:
really depends on your data, how fast you write the data in (gb per day) how often you want it to rotate and other considerations
there are some recommendation regarding bucket sizes for better search performance, but that also depends on some other factors