Well, some of the old numbers (Batting average, home runs, stolen bases and ERA, specifically) do an alright job, I guess. Imagine a line that measures a player's "true value" to his ballclub. A specific player would be a specific point on that line; whether he's good or bad or whatever, he's a point on that line, and we want to find out EXACTLY where he lies on that line. (Let's say, for the sake of argument, that we want "value" in terms of a player's ability to create runs through things only he can control, either directly so on offense or indirectly so on the mound)
Batting average, ERA and the others will give you a range on that line, meaning they will give you an area on that line where the player's worth resides, but they won't get you his exact position. This is due to the vagaries of things like AVG (a "hit" isn't always a real hit, and it also relies somewhat heavily on things outside of the player's control: the pitcher, the park, the defense, luck, etc; all this on top of not having as high of a correlation with run production--and year-to-year repeatability--as other metrics), RBI's (the player has no control over whether players before him were able to get on-base, therefore he should not be credited over a player with similar production levels that just happens to play on a team that can't get on base for him) and ERA ("earned runs" aren't always "earned" by the pitcher, as the defense might not be able to get to batted balls in the first place, the pitcher might get unlucky and hey, what about the ballpark?).
Things like OBP, SLG, ISO, OPS, wOBA, wRC, FIP, SO/BB, WHIP (coupled with BABIP), on the other hand, all have high correlations both in terms of being able to be repeated from season to season, as well as run production or prevention, whatever the case may be. Many of these things, while they may not account for or remove fully the reliance on defense and luck, do a much better job of isolating things a player has direct influence over, and therefore they get us much closer to that player's point on the line than the other, more dependent stats.
I guess things like AVG and ERA shouldn't be abandoned altogether (nor will they ever, probably), but to use them to value and credit players over the other, more descriptive metrics is just plain silly. That being said, things like RBI and pitcher W-L tell us next to nothing about the players we view through each respective lens, and have no business being involved in any baseball discussion ever.
