Shelton was a decent analytics pick because people thought he could become good against the pass. He is slow as dirt but strong, and because of that, scouts and geeks thought that he would develop the ability to break through the offensive line.
He did not. Nonetheless, during his first few years in Cleveland, he was good against the run. Two problems:
First, starting a guy because they are good solely against the run does not make much sense. The difference in rush yards per attempt will never be greater than -2, and when most good RBs average ~4 ypc, it’s just not enough difference to sacrifice pass defense.
Second, Shelton is not even as good as he once was against the run. That’s a problem because, even if he gives a -.3 yards against the run better than anyone we have, that’s not nearly enough to make up for his inefficiency against the pass.