Here is the Wikipedia article for John Hollinger's player efficiency rating (PER) statistic, which is commonly used as an "all-in-one" for (mostly offensive) performance in basketball. I would like to know how Hollinger cooked this formula up. I would guess it was based on some kind of historical regression, but I haven't been able to find out exactly (or even vaguely) what it is anywhere on the internet.
Yes it came from a sort of regression on statistics that are most important in "winning" a basketball game historically. He used the coefficients for certain counting stats (rebounds, assists, etc.) as the weights on these terms -- roughly. He also adjusted for things like minutes, pace, and the rest of the league.
One of the arguments against Hollinger's PER and some other ESPN stats are that there is not complete transparency in the origin of the statistic. Thus, something like adjusted real plus/minus is a great statistic but without a complete knowledge of how it is constructed it is difficult for us to truly value.