1/6 is not that small of a sample. Obviously those nine games mean absolutely nothing in terms of the type of player Parra is, but they do matter in the sense that the already very thin marginal gain from Snider to Parra is that much thinner. Parra is pretty clearly a better player than Snider, but he's not so much better that he's guaranteed to outplay Snider over the course of two months. In fact, there is a chance that Snider would outplay Parra, even if Parra is the better player.
I'm pulling these numbers out of my netherregions, but let's just say 75/25 that Parra outplays Snider to Snider outplays Parra over the period of two months. That's probably pretty generous to Parra, but whatever. Now, what about the odds that they play similar enough to eachother that there is no effect on the Orioles win/loss record. I'd put that at at least 40%, considering how close their career numbers are. That leaves only 35% chance that Parra makes any sort of difference. That's a pretty narrow margin, a margin we paid Davies for, and with 1/6 of his games already played for the Orioles, that margin has become even more narrow. We can quibble about all of those percentages--obviously they are not dead on--but I'm just using them to roughly illustrate how those 9 games of terrible numbers do matter in terms of the odds that Parra showing he was an improvement over Snider, let alone enough of an improvement to make any difference/justify the cost of Davies.