The only reason that many of these stats go from crappy to useful or even pretty good? Sample size. If you have 100 or 200 innings? ERA or RA9 starts becoming useful. Give me a thousand innings, and RA9 is pretty good. BABIP? Well, after several years, a pitcher’s BABIP is a pretty useful metric. A career’s worth of pitcher wins? Well, that’s useful too.

And therein lies the reason that people are fooled: because 300+ win level were all reached by great pitchers, then the pitcher Win stat MUST represent greatness, even at the seasonal level. So, even though the significance of a metric is dependent on the sample size, once that significance is reached, the reason for its significance (sample size) is summarily ignored.