Once again, to recap our overall philosophy: PRODUCTION + PLATE SKILLS + AGE-ARC
We got a new stat to measure "Production," which is called "Plausibility Index," which is covered via the "Allegory of the Window." I felt there needed to be a stat that incorporated both walks and extra-base hits, but also recognized that strikeouts impact Production by reducing the number of balls in play, and, thereby, increasing the necessary conversion rate (the rate of conversion of balls in play to "random-y singles") to a level that may (if it's very high) make offensive success "implausible."
So what about "Plate Skills," the other part of the formula?
Plate Skills can be measured by on-base percentage, or by "eye ratio" (BB / K), or by separate calculations of K% (K / PA) and BB% (BB / PA). But, here again, I didn't think we were getting the whole picture. Just like I view strikeouts as a non-trivial part of Production, I view "hitting the ball with authority" as an integral part of Plate Skills.
In other words, it's not just a matter of distinguishing balls from strikes, it's distinguishing a "hitter's pitch" from a "pitcher's pitch." The latter two may be balls or strikes, so "just" strike-zone judgment is not enough. When the hitter gets a "hitter's pitch," what is he supposed to do with it? Hit it! So, except as it demonstrates the relative ability to avoid strikeouts (by connecting with the ball), "eye ratio" doesn't really cover that part of Plate Skills.
OBP, then, might be better ... except, (1) it doesn't tell you anything about strikeouts, and (2) it can be mightily affected by our old friend the "random-y single." As we've noted, the "random-y single" represents a ball hit without authority, and our system treats "random-y singles" as no better than "random-y ball-in-play outs." Our theory is that the ability to hit "random-y singles" against minor-league pitching doesn't really tell us anything about a prospects likelihood of major-league success.
So, once again, we devised our own stat, which we dubbed "Hitter's +/-"
Without going into excruciating detail (well, maybe we already have), we took what appeared to be a reasonable "average" distribution of our six measurable "plate outcomes" from a bunch of major and minor leagues. [Again -- going back to the Manifesto -- we weren't looking for a "perfect" model, just something "reasonable."] [The six are walks, strikeouts, home runs, balls hit with authority (2b + 3b), singles (assumed "random-y" per our assumption), and ball-in-play outs (also assumed "random-y").]
Once we had this "normal distribution" of plate outcomes, there is an "expected" OBP resulting therefrom. It was .295. If a hitter's results reflected that "normal distribution," he would have a "Hitters +/-" of 0.00, because he would have neither increased nor decreased his "expected" OBP.
OK (maybe?) ...
We then take each plate appearance and measure the difference between that particular plate outcome and the "expected" OBP. A strikeout drops that PA's expected OBP from .295 to .000, so each strikeout is weighted at -.295. A home run or a walk increases the "expected" OBP from .295 to 1.000, so they are weighted at +.705. Balls in play hit with authority (measured as doubles and triples) we calculated (somewhat arbitrarily) to be worth +.225.
Singles and balls in play hit without authority we assume are neither positive nor negative (but "random-y"). Therefore, they neither increase nor decrease the "expected" OBP. In other words, they don't count in this equation at all.
So ... for a hitter to achieve a positive "+/-" the value of his weighted XBH + BB must exceed the value of his weighted K. Singles and ball-in-play outs are assumed out of the equation. If a hitter achieves that positive value, then he is turning the plate appearance to his advantage vis-a-vis the pitcher.
That last thing, ultimately, is what we are driving at. The hitter must "play defense" against the pitcher's attack in order to "play offense" against the other team. The ability to do that is what this stat is driving at.