Bases Empty: .251/.312/.396, .707
Runners On: .261/.331/.402, .733
There's also the simple matter of looking at 158 AB vs. 116. If you're trying to predict a player's true skill from that sample, the confidence intervals are massive.
Basically, it's something like this:
His OPS with the bases empty is something between .637 and 1.037. His OPS with runners on .658 and 1.158. We can't say with any certainty within those windows what his "true talent" split is (in this case, true talent could be defined as including Billy Hamilton -- this is about variance given samples).
Given that, how confident would be you that the observed split is representative of the actual one? We just can't read much of anything in to an observed split like that. And that's what Doug was saying.